Journal of Classification 1:147-186(1984)
lournal of
Classification
©1984 Sprmger-VerlagNew York Inc.
GENFOLD2: A Set of Models and Algorithms for the GENeral UnFOLDing Analysis of Preference/Dominance Data Wayne S. DeSarbo
Vithala R. Rao
University of Pennsylvania
Cornell University
Abstract: A general set of multidimensional unfolding models and algorithms is presented to analyze preference or dominance data. This class of models termed GENFOLD2 (GENeral UnFOLDing Analysis-Version 2) allows one to perform internal or external analysis, constrained or unconstrained analysis, conditional or unconditional analysis, metric or nonmetric analysis, while providing the flexibility of specifying and/or testing a variety of different types of unfolding-type preference models mentioned in the literature including Carroll's (1972, 1980) simple, weighted, and general unfolding analysis. An alternating weighted least-squares algorithm is utilized and discussed in terms of preventing degenerate solutions in the estimation of the specified parameters. Finally, two applications of this new method are discussed concerning preference data for ten brands of pain relievers and twelve models of residential communication devices. Keywords: Multidimensional scaling; Unfolding analysis; Preference models
1. Introduction Carroll and Arabie (1980) present an extensive taxonomy of measurement data and multidimensional measurement models, updating the work of Coombs (1964). Their schema attempts separately to classify data and models. In classifying data, these authors discuss seven distinguishing properties: number of modes, power of a given mode, number of ways, scale of data, conditionality of data, completeness of data, and number and nature of replications. For their taxonomy of models, they list four relevant characteristics: type of geometric model, number of sets of points in space, Authors' Addresses: Wayne S. DeSarbo, Marketing Department, Wharton School, University of Pennsylvania, Philadelphia, PA 19104 and Vithala R. Rao, Graduate School of Management, CorneU University, Ithica, NY 14853.
148
W.S. DeSarbo and V.R. Rao
number of spaces or structures, and degree of external constraint on model parameters. The authors then attempt to classify existing models and methods on the basis of these criteria. We wish to present a general class of models: GENFOLD2, GENeral UnFOLDing Analysis-Version 2, which, according to the Carroll and Arable (1980) classification, analyzes two-mode, polyadic, two-way, ratio or interval or ordinal scale, unconditional or conditional (assumptions concerning the comparability of the data), complete data. GENFOLD2, like traditional unfolding models, is a spatial, distance model, positioning two sets of points in the same space, allowing for constrained or unconstrained solutions. GENFOLD2 is an improved, modified version of GENFOLD (DeSarbo and Rao 1983), utilizing a more efficient algorithm and, as shall be discussed, provides joint space solutions which are "nondegenerate." We first describe the analytical problem of unfolding and review the relevant literature. GENFOLD2 is then presented in detail and the algorithm employed in the estimation of the parameters is discussed. Two applications of GENFOLD2 to preference judgments for over-the-counter pain relievers and intention to purchase statements for twelve types of new residential communication devices are discussed in detail, as are the managerial implications from a marketing point of view. Finally, some directions for future research are discussed.
2. Literature Review A number of different models and algorithms for estimating them have been proposed to account for individual differences in preferences. Here, the data are typically two-mode, polyadic, and two-way, where subjects render preference judgments on a specified set of stimuli. The psychometric literature on preference models has focused on two distinct types of models -- Tucker's (1960) vector model and Coombs' (1964) unfolding model Both models assume that subjects arrive at their preference judgments by considering a multidimensional set of stimulus characteristics, but differ in assumptions about how subjects combine stimulus information to arrive at a judgment. Davison (1976) and Carroll (1972, 1980) compare these two types of models and discuss the assumptions and implications of each. Focussing upon the unfolding-type, distance, spatial models, Bennett and Hays (1960) first generalized Coombs' (1950) unidimensional unfolding model to the multidimensional case. Here, both subjects and stimuli are represented as points in the same multidimensional space. The points for individuals represent "ideal" stimuli, or optimal sets of stimulus values (e.g. objects, brands, products, etc.) for those individuals. In the Bennett and Hays model, the farther a given stimulus point is from an individual's ideal point, the less the individual likes that stimulus. This notion of
GENFOLD2 for Analysisof Preference/DominanceData
149
relative distance implies a metric on the space. Bennett and Hays assumed this to be Euclidean -- implying that the isopreference contours, in two dimensions, are a family of concentric circles centered at the individual's ideal point (hyperspheres in higher dimensions). Several authors have proposed algorithms for estimating stimulus scale values and ideal point coordinates from preference judgments assumed to follow the unfolding model (Lingoes 1972, 1973; Bennett and Hays 1960; Roskam 1973; Young and Torgerson 1967; Kruskal et al. 1973, 1977; Kruskal and Carroll 1969; Sch/Snemann 1970; Carroll 1972, 1980; Takane, Young, and de Leeuw 1977; Heiser 1981; Spence 1979; Greenacre and Browne 1982). This approach of estimating both ideal points and stimuli coordinates is known as internal analysis (Carroll 1972), as opposed to external analysis methods which estimate only ideal points given the stimuli coordinates (obtained from perhaps an MDS analysis of similarities). Carroll (1972, 1980) has introduced PREFMAP and PREFMAP2 as a series of models and algorithms to perform analyses of preference data. His methods allow the user to select between internal or external, metric or nonmetric, and unfolding or vector model analyses. Three different (nested) unfolding models can be estimated in PREFMAP and PREFMAP2. Let: i
--
1 .......... I subjects;
j
=
1 .......... J stimuli;
t
--
1 .......... T dimensions.
Mq
---
the preference value subject i renders for stimulus j;
d~
=
the squared distance between subject i's ideal point and stimulus j ;
xjt
--
the t-th coordinate of stimulus j;
YJt
---
the t-th coordinate for subject i's ideal point;
w,t
=
the importance or salience weight of dimension t to subject i;
at
--
multiplicative constant for subject i;
bt
--
additive constant for subject i.
Then, the unfolding model (Carroll 1972) can be written in general form as:
W.S. DeSarbo and V.R. Rao
150
M q ~ F 7 1 (d2)-- a i d 2 + b i
,
(1)
with ai presumably negative; the specifications of the various unfolding models appearing in the literature can be treated as special cases of this general model. The simple unfolding model (Coombs 1964, pp. 140-180) defines di} as: T
d.2= ,j ~., (xjt - yit) 2
(2)
This "simple unfolding model" assumes that a given distance on a dimension makes as much difference to one subject as to another, as well as assuming that all individuals use the same set of dimensions within the space. Sch6nemann (1970) presents an analytic internal solution for a strong case (treating the data as unconditional) of this unfolding model. Ross and Cliff (1964) provide methods for obtaining a stimulus configuration through a singular value decomposition of the double centered M - ((M~j)) matrix under suitable assumptions. Sch~Snemann and Wang (1972) combine this metric unfolding model with the Bradley-Terry-Luce choice model (Luce 1959), to produce a stochastic unfolding approach that is applicable to paired comparisons data. Zinnes and Griggs (1974) present a probabilistic multidimensional analogue of this model. Davidson (1972, 1973) presents a geometrical analysis for this type of unfolding model. Carroll's weighted unfolding model defines di~ as: T
di~ = ~., wi,(xjt - Yit) 2
(3)
t=l
Here, subjects are allowed to weight the dimensions differently, where wit can be thought of as the "salience" or "importance" of the t-th dimension for subject i. Now the isopreference contours are ellipses, ellipsoids, or hypereUipsoids instead of circles, spheres, or hyperspheres, respectively, as in the simple unfolding case. Srinivasan and Shocker (1973) present a nonmetric external unfolding analysis with this model using linear programming methods including nonnegativity constraints for the dimension weights. The same constraints are provided in a metric procedure using quadratic programming described by Davison (1976). Spence (1979) presents an interesting generalization of this model allowing for linear constraints on x = llxi, ll and/or Y = lly,,ll. Finally, (1972, 1980) Carroll's general unfolding model specifies di} as: T
= Z t=l
y;)2
,
(4)
GENFOLD2 for Analysis of Preference/Dominance Data
151
where Xj = XjTi (Xj = the j-th row of X), Yi* = YiT; (Y~ = the i-th row of Y), and T~ = an orthogonal transformation matrix for the i-th subject. This unfolding model allows subjects the freedom of implicitly choosing idiosyncratic sets of reference axes within the space. That is, each subject is allowed to rotate the reference frame of the perceptual spaces and then to weight differentially the dimensions defined by this rotated reference frame. Thus, X~ depends on i. Bechtel (1976) also presents a related general powered distance model for unfolding. Ramsay (1982) and de Leeuw and Heiser (I979) also consider such metrics for dissimilarity data. There is controversy in the literature over the desirability of constraining wi,'s in the weighted (and general) unfolding model to be positive. Carroll (1972) claims that in the weighted unfolding model, a negative wit has a clear interpretation -- if vet is negative with respect to dimension t, the ideal point for individual i indicates the least preferred, rather than the most preferred, value and the farther a stimulus is along that dimension from the ideal point, the more highly preferred is that stimulus. He thus argues for not constraining the w,t to be positive. Other authors such as Srinivasan and Shocker (1973) and Davison (1976) dispute the value of unconstrained analyses, claiming that the existence of such negative weights may lead to unrealistic predictions for subjects' most preferred stimuli.
3. GENFOLD2
3.1 Objectives Our primary goal is to provide a general, flexible unfolding model which can accommodate: conditional or unconditional analyses, internal or external analyses, W = II w,t ll constrained or unconstrained analyses, ratio, interval, or ordinal scaled data, and linear constraints on Y or X. This last feature proves to be of considerable utility in many types of applications, as we discuss later in the paper. In addition, we wish to create an unfolding procedure which will not render degenerate solutions which have plagued many past efforts to fit such models. Thus, GENFOLD2 allows the user to fit a large class of different types of unfolding models, combining the features of many types of unfolding models encountered in the literature.
3.2 The Model Let: Aij
The "dispreference data value" (inversely related to preference data values) the i-th subject has for the j-th stimulus;
152
W.S. DeSarbo and V.R. Rao
Y~t
=
the t-th coordinate of subject i's ideal point;
xjt
=
the t-th coordinate of stimulus j ;
Wj
=
subject i's linear (symmetric) transformation matrix;
a~
ffi
subject i's multiplicative constant;
bi
--
subject i's additive constant;
c~2
--
subject i's exponent;
f~j
ffi
squared distance between subject i and stimulus j ;
ee
--
error;
Aa
=
the l-th descriptor variable for subject i;
art
--
the importance or impact of the l-th descriptor variable on dimension t;
Bjk
=
the k-th descriptor variable for stimulus j ;
Okt
=
the importance or impact of the k-th descriptor variable on dimension t;
i
=
1 .... , I
subjects;
J
=
1 .... , J
stimuli;
t
=
1,...,T
dimensions;
!
--
1,...,L
subject descriptor variables;
k
=
1,...,K
stimulus descriptor variables.
Then, the full G E N F O L D 2 model can be written as: Au = / ~ u + eu where:
~kij= and
fc~2 + bi
ai.~ ij
,
(5)
153
GENFOLD2 for Analysisof Preference/Dominance Data
=
-
y,),
3.2.1 Conditional vs. Unconditional Analyses GENFOLD2 allows a variety of options for the estimation of ai, b~, and q parameters. These parameters prove to be very important sources of information concerning the structure of the data used in the analysis. For example, the a~'s in the simple unfolding context, when negative above, indicate the presence of an "anti-ideal" point, similar to the interpretation of negative weights in the weighted unfolding model. The exponents, c7, which are constrained purposely to be nonnegative for interpretability, render information as to the rate that preference decreases as one moves away in any direction from subject i's ideal point. Finally, the b,.'s are additive constants, appropriate for interval scale data. In the fully conditional case, all three sets of parameters are unconstrained. In the fully unconditional case, the user has the options of setting a~ -- a or some fixed constant (e.g., a = 1), b~ = b or some fixed constant (e.g., b = 0), and for c2 = c or some fixed constant (e.g., c = 1). The user is allowed the flexibility of mixing both constrained and unconstrained sets of parameters with respect to individual subjects. 3.2.2 Internal vs External Analysis Assuming that a configuration X has been derived (call it X*) from a previous analysis (e.g., from an MDS analysis of similarities), one can constrain X = X* in the analysis (external analysis). Or, one can estimate X (internal analysis) from the data. • 3.2.3 Form of W ~ One set of constraints on W i gives the user the flexibility of specifying a particular type of unfolding model. For example, with q2 = c = 1, Vi = 1 . . . . . 1, then by constraining W'----I (a Tx T identity matrix), V i= 1 . . . . . 1, one obtains Carroll's simple unfolding model. By constraining W i to be diagonal, V i = 1. . . . . I, one obtains Carroll's weighted unfolding model. Similarly with ci 2 = c = 1 , V i = 1 . . . . . I , and W i = UiU '', one obtains Carroll's general unfolding model. 3.2.4 Ratio, Interval, or Ordinal Scale GENFOLD2 is designed to handle all three of these types of data. For ratio scale data, bi -- b -- 0, ~/i-- 1 . . . . . I, while for interval scale data, the additive constant bi is estimated or constrained to some b (not necessarily equal to zero). For ordinal scale data, as to be discussed, a monotone
154
W.S. DeSarbo and V.R. Rao
regression phase (Kruskal 1964b) is appended to the metric GENFOLD2 algorithm. 3.2.5 X and/or Y Constrained vs Unconstrained Finally, GENFOLD2 allows one to constrain either Y or X or both to be linear function of some known set of preselected descriptor background variables such as demographic or product feature variables (A for subjects, B for stimuli), i.e., Y - A/a X-- BO
(6)
As in CANDELINC (Carroll, Pruzansky and Kruskal 1980) and in threeway multivariate conjoint analysis (DeSarbo, Carroll, Lehmann, and O'Shaughnessy 1981), these constraints can aid in the interpretation of the dimensions derived (c.f., Bentler and Weeks 1978; Bloxom 1978; Noma and Johnson 1977; de Leeuw and Heiser 1978; Lingoes 1980) and can replace the property-fitting methods often used to attempt to interpret results after a solution is obtained. Also, as shall be discussed shortly, the imposition of these sets of constraints can provide an effective tool for product positioning and market segmentation analysis.
3.3 The Algorithm In essence, GENFOLD2 attempts to estimate the desired set of constrained and/or unconstrained parameters described (i.e., some subset of: W j, X, Y, a, b, c, a , 0) given A and T (the number of dimensions) in order to minimize the weightedsum of squares objective function:
Min Z = E E ?u [Au - Ai/12 i
(7)
j
There has been considerable research attempting to cure unfolding of its tendency toward degenerate solutions. Degenerate solutions often occur in multidimensional unfolding in a number of ways. One common degenerate pattern that frequently occurs (in two dimensions) is where the two sets of points lie in concentrentric circles, where one set of points encircles another. Another typical degenerate solution often encountered is one where (in two dimensions) the two sets of points are separated from one another on one or more of the dimensions (e.g., the ideal points all in quadrant one and the stimuli all in quadrant three of a two-dimensional space). Such degeneracies have been investigated by Kruskal and Carroll (1969)
GENFOLD2 for Analysisof Preference/DominanceData
155
who propose different "badness-of fit" measures to solve this problem. Their method tends to work nicely with artificial data, but according to Carroll (1972), their technique "has so far performed less than fully satisfactory with real data." Carroll (1972, 1980) also acknowledges this degeneracy problem associated with the unfolding model. Heiser (1981) imposes configuration restrictions in the two-way case on the X = t txj, ll and re = I lyt, II joint space. The restriction Heiser imposes is that each stimulus is constrained to be at the centroid of the location of the subjects (ideal points) for whom it was most preferred (first chosen), or, more generally, among the first M ~< J most preferred stimuli. While this strategy is clearly a major contribution, several limitations arise with such an approach. One limitation with the Heiser approach concerns the assumption regarding the reliability of the data. Basically, Heiser's (1981) approach tends to place most of the weight in estimating parameters on the first M choices (most preferred stimuli), tending to down-weight the rest of the data. This strategy is roughly equivalent to assuming that a subject can only reliably render preference information about a brand most preferred, or the first M preferred, while the other judgments may be too "noisy" to carry significant weight in estimating model parameters. While this "implicit weighting scheme" may be viable for some particular applications, it is certainly not realistic for many. Depending upon the nature of the study, questionnaire design, number of stimuli, discriminability between stimuli, type of subjects, knowledge of the stimuli by subjects, etc., different assumptions may be appropriate concerning how much error preference judgments contain. For example, in product testing for an automobile, a consumer may be able to render reliable judgments on the first M preferred automobiles. Or, it may be the case that the typical subject could provide reliable ratings for a favorite M1 cars and for a least favored M2 cars. That is, it is conceivable that a subject can reliably tell the interviewer what he/she likes and dislikes, but may have trouble rating cars in between. Such data may require a quite different weighting scheme, if the aim is to weight more heavily the more reliable responses. We propose a different approach to the degeneracy problem in unfolding, similar to that of DeSarbo and Carroll (1983). Our approach involves explicitly altering the loss function in expression (7) to incorporate data weights 3'u, where the 7u are defined by the user to weight the A U values differently. We share Heiser's (1981) implicit theory about a possible cause for degeneracy being the error or noise in the data, and we thus provide the flexibility of the user specifying 7;j differently. For example, one may define the weights as: 7ii
=
(8)
W.S. DeSarbo and V.R. Rao
156
where p is some exponent. Assuming the Aij are dispreference data values, such a weighting scheme would weight smaller values (more preferred) higher than larger values (less preferred). In cases where specific preprocessing is involved (e.g., subtracting means) and negative values of A/J appear as interval scale data, such a weighting function may not be meaningful. Here, for example, one could use:
"F/j----
(9)
where p is an exponent and r(A 0) represents the row ranks (from smallest = 1 to largest = J) of the A/J. Note, as p-'-' ~ , this (row) weighting scheme in (9) resembles Heiser's (1981) configuration restrictions (M ~ 1) since only first choices would be significantly weighted (as 'Fu = 1 for first choices), while the rest of the 3,/j would tend to zero. (It should be noted that this limiting case would not be precisely equivalent to Heiser's approach.) Other weighting options are also possible. For example, one could specify 'F/J = 1, Vi,j, so that the "weighted" loss function reduces to the nonweighted one. Or, one could specify a bimodal or step weighting function where, say, the first three and last three choices would be highly weighted, and all others received low weights. The choice of the "appropriate" weighting function depends upon such factors as the preprocessing options and scale assumptions of the data, the assumptions of the conditionality of the data, the assumptions concerning the reliability of the different data values, and, trial and error. As mentioned, if the data were treated as interval scale and preprocessed accordingly (e.g., by taking out particular row, and/or column means), then specifying 'Fu as in (8) may not make sense. Depending upon the conditionality assumption made for the data, it may or may not make sense to define "F/J via expression (9) over rows. Also, different "F/j could be specified depending upon the assumptions made concerning the reliability of the A U collected. As stated previously, if the researcher believes that only highly preferred judgments are reliably given, one could specify a unimodal 'F/J function. On the other hand, if one believes that the A/J can be reliably given for both highly preferred and highly non-preferred judgments, then a bimodal weighting function can be specified. While the first three criteria may provide insight into the general form of "F/J (e.g., unimodal vs. bimodal, or expression (8) vs. expression (9)), specific decisions, such as what value p should have, can be made by trial and error, although as will be demonstrated in the example to follow, in expression (9), p - - 2 appears to work well. We believe that the
GENFOLD2 for Analysisof Preference/DominanceData
157
specification of p is an empirical issue depending upon the data. One approach is to run the analyses with a different sequence of p's (p = 0,1,2,3,4) and examine at what point degenerate solutions disappear. Another approach worthy of further research would be to estimate p while fitting the model. There are some twelve steps involved in the GENFOLD2 algorithm. Each step is described in turn. 3.3.1 Input In addition to the two-way, two-mode data matrix of dispreference values A, the user must input various control parameter values designating: the number of dimensions (T) for the analysis, preprocessing options, method of generating starting values, type of unfolding model, type of data scale, type of analysis (external or internal), whether Y is constrained, whether X is constrained, desired c~ option, desired bi option, desired ai option, W ~ constraints, maximum number of major iterations (MAJOR), maximum number of minor iterations (MINOR), convergence tolerance (TOL), maximum number of iterations for nonmetric procedure (NMET), and, type of weighting function. These input control values will be described in detail in the discussion of the main parts of the algorithm to follow. 3.3.2 Preprocessing GENFOLD2 allows the user to either use the raw data or to utilize one of a variety of preprocessing options such as: row center A, row center and row standardize A, row and column center A, double center & and row standardize, remove geometric mean from rows, remove geometric mean from columns, normalize columns to unit sums of squares, and normalize rows to unit sums of squares. Clearly, the choice of preprocessing option will depend on such factors as the assumption concerning the scale of the data, the units of measurement, conditionality assumptions, method of data collection, etc. 3.3.3 Generate Starting Values GENFOLD2 has been programmed to allow for a variety of different starting values for W i, X, Y, etc. One can choose from the following starting procedures:
3.3.3.1 Random Start Here, random and feasible values are generated for W', X (or #), and Y (or a ) from a uniform distribution. The values, especially for W j, must be feasible (e.g., diagonal if W i is constrained to be so)
158
W,S. DeSarbo and V.R, Rao
for the algorithms to function properly as will be demonstrated. Also, at = 1 b~ -- O, and ct = 1, V i -- 1. . . . . I, in this starting option.
3.3.3.2 External Analysis. Here, X = X" is given having been derived through a scaling analysis of similarities or from some other method (e.g., from a hypothesis or theory). The X* is fixed throughout the analysis. Values for the remaining parameters are generated randomly as above. 3.3.3.3 Values GiveJt GENFOLD2 also allows one to start the analysis with predetermined values of all of the parameters. This proves to be a valuable option when testing the goodness-of-fit and stability of some hypothesized solution or one obtained from some previous analysis. In this option, GENFOLD2 immediately proceeds to the estimation phases of the algorithm. 3.3.3.4 A "Close" Start On X. Using the procedure demonstrated by Ross and Cliff (1964), Sch6nemann (1970), and Carroll (1980), one can double center (or single center) A and perform a singular value decomposition on this double (single) centered matrix: A - - PDQ'
,
(10)
and let the starting value of X be QD. Starting values for the remaining parameters are generated randomly as described above. Under certain restrictive assumptions demonstrated by Ross and Cliff (1964) and Sch~nemann (1970), this method of internally generating an X configuration will be "correct" up to an affine transformation. Note, one could also perform such a decomposition on the matrix t l A u • in order to take into account the specified weighting function.
3.3.3.5 Close Values for Parameters. Assuming c~ = 1, Vi = 1. . . . . I, "close" values can be defined for the simple, weighted, and general unfolding models using regression techniques adapted from Carroll's (1980) PREFMAP2. 3.3.3.6 Some General Considerations. Each of these three "close starting" algorithms resemble the initial procedure utilized in PREFMAP and PREFMAP2 to derive X, Y, W ~ estimates. For the simple and weighted unfolding models where a rotational problem exists, the PREFMAP and PREFMAP2 procedure then, seeks to find an appropriate transformation (for orientation and/or weighting) for the coordinate system. In the weighted unfolding model, only the orientation of axes needs to be determined since the weighting is determined separately for each subject. For simple unfolding, however, both orientation and weighting are solved for. (However, the
GENFOLD2 for Analysisof Preference/DominanceData
159
transformation solved for in PREFMAP and PREFMAP2 is not optimal in a least-squares sense.) One interesting approach would be to use the final PREFMAP or PREFMAP2 solutions as starting values for GENFOLD2 -work that is currently underway. Note that these methods of generating "close" starting values stated above ignore the weighting function Yu. Current efforts are progressing to have these weights built into the estimation procedure as an option. This would be done by simply multiplying the corresponding data in each case by 7 i~ and adjusting the estimates accordingly. Also, note that starting estimates for a and/or # can also be obtained after providing such starting values for X and Y. One can obtain starting values for a by regressing Y on A, and similarly obtain starting values for 0 by regressing X on B. 3.3.4 Estimate W ~ There are a variety of options in GENFOLD2 for imposing particular constraints on W t. For example, if a version of the simple unfolding model were desired, W ~= I, an identity matrix, and the analysis would skip this phase of the algorithm for all subsequent iterations. Other major constraint options concern particular combinations on whether W j = UiU e is to be diagonal or generally symmetric. Note, by requiring that W " = U ' U e, we force W t to be symmetric and at least positive semi-definite. Both make sense given the quadratic form of the squared distance f e given in equation
(5).
3.3.4.1 General s y m m e t r i c W i. For this selection, a general, unconstrained symmetric I x Tx T array (W) is to be estimated. Since one can express W i = UiU ~', we calculate partial derivatives of the loss function in equation (7) with respect to Ui: 0_.__z.z= _ 2 E
OUi
0£j
'u(au -
(11)
a ^
c?--i
= - 2a, ci2 ~_~ T,j(A e - A i j ) f e
(Xj
--
Yi)'(Xj
-
Yi)U i
(12)
These partial derivatives are to be used in a Quasi-Newton unconstrained algorithm to provide an approximation to the Hessian matrix rendering information of the curvature of the objective function. The algorithm used is a modification of the Davidon-Fletcher-Powell Method (Davidon, 1959; Fletcher and Powell 1963).
160
W.S. DeSarbo and V.R. Rao
Let: r
=
I T 2,
Hk
=
an r x r positive definite symmetric matrix, at the
hk
=
optimal step length at iteration k,
VZk
=
[ 0 ~ ] = the gradient of the objective function at iteration k,
Sk
----
the search direction at iteration k.
k-th
iteration;
The steps of the iterative algorithm used are as follows: 1.
Start with given values (U), and an r × r positive definite symmetric matrix H1 = I (identity matrix) initially. Set k -- 1.
2.
Compute V Z k at the point (U)k and set S k -- -
HkVZ
k
(13)
Note that for the first iteration, the search directions will be the same as the steepest descent direction -- - VZ1, providing H1 -- I. 3.
Find the optimal step length h k* in the direction Sk. This is done through use of a quadratic interpolation (see Cooper and Steinberg 1972) search procedure. Then we set: (U)k+I ---- (U)k + hk*Sk •
4.
(14)
This new solution (U)k+l is tested for optimality and for maximum number of minor iterations. That is, we see if." (a) ( Z k Zk+l) <~ TOL, or (b) k > MINOR. -
If either of these two conditions hold, this procedure is terminated. If neither hold, then we proceed to step (5). 5.
Update the H matrix as: H k + 1 ---- H k d" M k d- N k
where:
(15)
161
GENFOLD2 for Analysis of Preference/Dominance Data
I
SkSk
M k = h~
(16)
SkQk
'
(HkQk)(HkQk)'
Nk = --
t
,
(17)
QkHkQk
Qk 6.
(18)
VZk+l-- VZk
=
Set k = k + 1 and go the step (2).
Gill, Murray, and Wright (1981) provide a derivation of this procedure as well as its convergence properties. Note, we could have easily performed this same algorithm on W j instead of on U ~. However, letting W i = U;U e allows us to convert a constrained problem (symmetry and positive (semi) definiteness) into an easier unconstrained problem. The use of this QuasiNewton method had been favorably compared with other gradient search procedures such as steepest descent and conjugate gradient methods (Himmelblau 1972). It was found empirically that the approximate second derivative information can aid in speeding up convergence, especially when near the optimal solution. In addition, since the first step of this algorithm was a steepest descent search, one could take advantage of a steepest descent search when initially far away from the optimal solution (empirical research demonstrates that steepest descent is best used in early iterations when far from the optimal solution). 3.3.4.2 D i a g o n a l W i. The minimization problem to be solved here is to
minimize: I
J
z = E Z ru(a,j i=1 j = I
subject to: (i).
w/s > / 0
V i = 1.....
Vsfr=l
I .....
T
s = 1.....
T
and (ii).
w]s = 0
V r #
Vi = 1 . . . . .
(19)
I
While algorithms exist (e.g., Lawson and Hanson 1974) to solve such an equality and inequality constrained, nonlinear optimization problem using
162
W.S. DeSarbo and V.R. Rao
Kuhn Tucker conditions, the computational cost of such procedures becomes enormous for large IT. One alternative procedure is to re-express W i = UiU i' to enforce symmetry and nonnegativity. This decomposition is defined (see Green 1978) for all nonnegative definite symmetric matrices via the singular-value decomposition principle for symmetric matrices. Thus, we can convert the problem to one merely involving equality constraints which is much easier to solve. For this problem, a projected Quasi-Newton method is to be used. Let:
Ogl(h (k)) .....
Ogl(h (k)) .
.
Ohl
,
Ohb
(20)
A*=
Oga(h (k))
Oga(h (k))
0hi
Ohb
be the a x b matrix of partial derivative of the (active) equality constraints where: gs
ffi
The s-th equality constraint equation,
h (k)
=
the vector of parameters to be estimated whose elements are the b --- I T 2 elements of U
=
1. . . . .
IT(T-l)
=
1. . . . .
(IT 2) parameters,
=
the k-th iteration.
k
constraints,
Here, there are I T ( T - I ) equality or active constraints for the off-diagonal elements of U. One of the advantages of converting the problem into a pure linearly constrained one is that the matrix A* and the associated projection matrix: B*ffi I - A * ' ( A * A * ' ) - I A *
(21)
need only be calculated once in the entire program since all the constraints are linear equality constraints (no inequality constraints) and are therefore,
G E N F O L D 2 for Analysis o f P r e f e r e n c e / D o m i n a n c e Data
163
by definition, "active" or binding at every iteration in this phase of estimation. With B*, now computed, one can define the projected gradient as:
VZk"= B*VZk
(22)
and the projected Hessian approximate as: Hk*=
B*'HkB*.
(23)
These are then substituted in the algorithm above for V Z and H respectively, and the same general procedure is utilized. Gill, Murray, and Wright (1981) suggest alternative reduced space procedures to obtain B* which might prove more efficient. Current work is being performed in this area. 3.3.5 Estimate a and/or Y
3.3.5.1 Y Unconstrained. The optimization problem at this phase involves attempting to minimize the loss function in (7) with respect to Y. Taking partial derivatives:
O__z_z= _ 2 Z YJJ(ao- Au) -0A,: ~ OYi
j
= -- 4aici 2 ~
Tu(Aij
--
i j ,~J ri jc ) - l (.w - -i - ¥ i
' - - WiX~)
,
(24)
J
the unconstrained Quasi-Newton algorithm described in 3.3.4.1 is utilized to estimate Y.
3.3.5.2 Y = A a. Given an I x L ( L < I ) matrix of background descriptor information which presumedly contains relevant data to fully describe the behavior of Yit, we attempt to minimize the loss function in (7) with respect to a. Taking partial derivatives:
2 E E ~.(a. - L~) oL:
0z 0a
i
=
4 ~., ~ i
j
j
Tij(Au
Oa ~' "~ 2 ~ciZ - I ' " r_ _ i t -- zaij)aic i Jd (AiXjW -- AiAiaW
i
) ,
(25)
164
W.S. DeSarbo and V.R. Rao
where Ai is the i-th row of A, the unconstrained Quasi-Newton algorithm described in 3.3.4.1 is utilized to estimate a . Once ot is estimated, Y is updated via Y = Aa. 3.3.6 Estimate 0 and/or X 3.3.6.1 X U n c o n s t r a i n e d . Now, the optimization problem reduces to one involving attempting to minimize (7) with respect to X. Taking partial derivatives:
0 ~Z 0Xj = - 2 ~,, 7ij(Ai./_ AU) AUOXj _ .2tc?-ltwiv ' = - 4 ~., 7 i j ( A U - A U ) , i c i Jij "'" " j -
WiY;)
(26)
i
the unconstrained Quasi-Newton algorithm described in 3.3.4.1 is utilized to obtain an estimate of X. 3 . 3 . 6 . 2 X = B0. Given a J x K matrix of descriptor variables describing the
important aspects of the stimuli, we attempt to minimize the loss function in (7) with respect to 0. Taking partial derivatives:
i
j
= 4 ~ ~ Tu(A,j i
00 ~ •~ 2 t c i L 1 r i / U,,,;~, Ju (BjY/W -
r
i
BjB/OW )
,
(27)
j
where Bj is the j - t h row of B, the unconstrained Quasi-Newton algorithm described in 3.3.4.1 is utilized to obtain an estimate of O. Once an estimate of 0 is obtained, X is updated via X = B0. 3.3.7 Estimate ciz Assuming metric (ratio or interval scaled) data, several options exist, as discussed, concerning the estimation of c7. The simplest option is to constrain c/2 = c = 1, ~ i = 1 . . . . . I, in which case this phase is ignored. Two other major options exist. 3.3. 7.1 ca U n c o n s t r a i n e d . The optimization subproblem at this stage entails the estimation of ci, V i = 1 . . . . . I, so as to minimize the loss function in (7). Taking partial derivatives:
GENFOLD2 for Analysis of Preference/Dominance Data
e z = _ 2 ]~ ~,,:(a U - ~ . ) 0ci j
165
o~,,j 0c;
= - 4aic~ ~,, "fu(Au - ~- u , ,~ rc'2u In di: , J
(28)
the unconstrained Quasi-Newton method described in 3.3.4.1 is utilized iteratively to provide an estimate of q. 3.3.7.2 ciffie, a constant V i = ~ l , . . . ,I. Here we constrain c~ = c22 . . . . . ct2 = c 2, a constant which we attempt to solve for in order to minimize (7). Taking partial derivatives:
o z = _ 2 Z Z ~,,j(a;j - ,i,;? o/~,; 0c
i
:
Oc ^
c2
= - 4 ~ ~ 7ij(A U - Aij)a, cif,~ In ~ j , i j
(29)
the same unconstrained Quasi-Newton method is iterativety applied.
3.3.8 Estimate bl and/or ai Assuming metric data, there are nine possible options in G E N F O L D 2 concerning the estimation of combinations of additive and multiplicative constants. Note, only when W i--- I, the identity matrix, V i = 1 . . . . . 1, does it makes sense to estimate an at since its effect would be absorbed in a general or diagonal W j. 3.3.8.1 ai ffi 1, bi ~ 0,, Vi = 1 , . . . ,I. In this trivial case where there is no additive constant and all multiplicative constants equal one, there is obviously no need for estimating anything. 3.3.8.2 ai - 1, bi "= b, Vi =" 1 , . . . ,
I. Let:
mu
--
d lJ '2 ~ r~ [(X/-
rnij
=
m U x ~/ ~/],
M *
=
a vector (of length I J) stringing out mu~,
Yi)W t (Xj-
Yi)'] c'2,
W.S. DeSarbo and V.R. Rao
166
~,
---
a vector (of length H) stringing out Tu,
A
--
a vector (of length IJ) stringing out A U,
A*
=
A
N
--
[y~h,M*] ffi an (H) by 2 matrix where the first column consists of ~,~ and the second column is M*.
x
~,,h,
Then, unconstrained estimates for a and b that minimize the loss function in equation (7) can be solved for via: F ffi (bl ffi (N'N)-IN'A"
(30)
If a is not equal to one which in mostly every instance will be the case, then a constrained regression analysis is performed forcing a =-- 1. This is done first by defining:
[01 and estimating the constrained coefficients via (Johnston 1972):
[ab} -- r + t(N'N)-IR'tR(N'N)-IR']-I(r - RF)],
(31)
which now constrains a -- 1. 3.3.3.3 ai - 1, bi unconstrained, Vi - 1 , . . . , Mi*
_-
I. Let:
a vector of length J for subject i containing elements mij, j = 1 . . . . . J, a vector of length J containing the preference values for J stimuli for the i-th person,
H i
~--
[T~h, Mi*] = a J by 2 matrix for the i-th subject where the first column consists of T ~h and the second column is Mi*.
Then, for each subject i, i - - 1 . . . . . I, unconstrained estimates for ai and bt that minimize the loss function (7) are obtained via:
G E N F O L D 2 for Analysis of P r e f e r e n c e / D o m i n a n c e Data
F~ =
bi ' 1 ' ai -- (NiNi)- N i A i
167
*
(32)
•
If a~ ;~ 1, then a constrained regression analysis, for each subject where a~ ;~ 1, is performed constraining a~ -----1. This is done in the same manner as in (31) where Ni and F~ replace N and F respectively. 3.3.8.4 al = a, bl = 0, V i = 1 , . . . , I. Using the notation above, we can estimate a multiplicative constant for the entire sample of subjects, assuming no additive constant, that minimizes the loss function (7) via: a = (M*'M*)-~M*'A*
(33)
without estimating an intercept term. 3.3.8.5 a i ~= a, bi = b, V i =, 1 , . . . , (7) can be obtained via:
I. Estimates of a and b that minimize
bI = (N'N)-~N'A*.
(34)
3.3.8.6 ai ~ a, bi unconstrained, V i -- 1 , . . . , I. Estimation of a and bg that minimize (7) is performed in two stages. In stage 1, a is estimated for the entire sample using equation (33). One estimated, it is embedded into the W~'s via simple multiplication and reset to 1. Then, a constrained regression analysis is performed to estimate bi with a,.--1 via the technique discussed above. Once this is done, the a is extracted from the W"s via division and a i is set equal to a, redefining the Wi's. 3.3.8.7 ai~ constrained, bi-- 0,V i ~ 1 , . . . , which minimize (7) can be obtained via: ai =
I.
Here, estimates of ai
(M,*' Mi*) -1 M,*' Ai*,
(35)
e.g., performing I separate regressions without an intercept term. 3.3.8.8 ai is unconstrained, bi = b,Vi = 1 , . . . , I. Estimation of at and b which minimize (7) is performed in two steps. In step 1, a, is estimated via equation (33) without an intercept term. In step 2, the aj's are absorbed into the W~'s, and a constrained regression analysis is performed for the entire sample to estimate b via equation (31) constraining a = 1. The ai's are then extracted from the W~'s (redefining the W"s).
168
W.S. DeSarbo and V.R. Rao
3.3.8.9 ai is unconstrained, bi is unconstrained. Estimates of ai and bi which minimize (7) are obtained by performing I separate regressions of the form:
{]ajbi ---- (N~ Ni) -1 Ni''
•i
"
(36)
3.3.9 Termination Criteria The algorithm terminates when either one of two conditions hold: a.
b.
The number of major iterations exceeds MAJOR; or, Iz
Concerning condition b above, one can show that the algorithm described will converge to at least a local optimum using a limiting sums argument (Courant 1965). 3.3.10 Non-Metric Analysis Assuming the data are on an ordinal scale, a nonmetric option to analyze such rank orders is provided. This step of the estimation phase entails performing a monotone regression analysis (see Kruskal, 1964a,b) for each subject in order to estimate a subject-specific monotone function (by subject). Since a general monotone function would cover multiplicative (positive), additive, and exponential transformations, steps 3.3.7 and 3.3.8 are not needed here. Also, the appropriate loss function is a STRESS measure (STRESS FORM-2 is used). This method is similar to that used in PREFMAP2, where the alogorithm alternates between the metric analysis and an MFIT procedure. 3.3.11 Test to Stop Termination would occur if either: I. 2.
Convergence of STRESS (i.e.,[STRESS k - STRESSk-II < TOL); or, M a x i m u m number of iterations are exceeded ( M I T ) N M E T ) .
If neither condition holds, then the redefined (monotonic) preference values are utilized in G E N F O L D 2 as the program returns to step 3.3.4.
3.3.12 Output The values of all estimated parameters are printed and plotted separately for the user. The overall goodness-of-fit measure (sums-of-
GENFOLD2 for Analysisof Preference/DominanceData
169
squares accounted-for - for ratio scale data, variance accounted-for -- for interval scale data, stress -- for ordinal scale data) is printed also. 3.3.13 Diagnostics GENFOLD2 contains a number of diagnostic aids to assess the quality of the fit.
3.3.13.1 Correlations Between Dimensions. One can assess how interrelated dimensions extracted are by examining correlations within each set of parameters (X, e, Y, a , and W). Excessively high correlations may indicate that too many dimensions have ,been extracted. 3.3.13. 2 Goodness-of-Fit Decompositions. Appropriate goodness-of-fit measures can be obtained by subject and/or by stimuli. This provides insight into possible outlier subjects and stimuli which may be inappropriately included in the study or, at least, render quite different types of responses.
3.3.13.3 Residual Analyses. A list of actual, predicted, and residuals is printed at the end of each analysis together with the corresponding subject and stimulus identification codes. In addition, plots of actual vs predicted, residual vs predicted, and residual vs. actual are output for detecting outliers and nonrandom patterns in the residuals. GENFOLD2, as all unfolding models, has the potential of estimating quite a few parameters given a fixed I by J data set. The model degrees of freedom can be written as: H - [81 82 JT + 81 (1 - 8 2 ) K T
+ (1 - 8 3 ) L T +
q- 841T q- 851 + 86 I + 87 I]
83IT
,
where:
81 =
{01 ifexternalanalysis {01 if X = B 0 if internal analysis, 82 = if unconstrained,
83 =
01 if Y = Aa if unconstrained,
(37)
w.s. DeSarboand V.R. Rao
170
0
if W'---I, V i = l ..... I if W' diagonal "
84 =
i 85 =
__.__1_1 if Wi general symmetric
i 86=
/I
/I
if a , = l VI----1,...,I if a t ~ a ' if a, unconstrained " ,
if c , 2 = 1 V , f = l .... , I ti if b,-=0 Vi=l,...,I if c 7 - c2 " 87--- / I i f b , = b if b, unconstrained " . if c~2 unconstrained " , t
4. Applications 4.1 Pain Reliever Preference Analysis 4.1.1 Study Description A study was conducted among a sample of 61 MBA students enrolled in a Marketing Management Course at Fairleigh Dickinson University (DeSarbo and Rao 1983) to investigate overall preferences for ten brands of over-the-counter pain relievers: Anacin, Ascriptin, Bayer Aspirin, Bufferin, Cope, Datril, Excederin, Hudson Aspirin, Tylenol, and Vanquish. Unlike other data collected in this study (c.f. DeSarbo and Carroll 1983; Carroll, Clark, and DeSarbo 1983), subjects were not given any other information about these drugs concerning ingredients, price, availability, etc. It should also be noted that these data were collected before the "Tylenol scare," and were thus not affected by that unfortunate incident. Previous analyses utilizing spatial (DeSarbo and Carroll 1983) and nonspatial (Carroll, Clark, and DeSarbo 1983) models have been performed on three-way data (subjects rendering intention to use ratings for these brands in relieving three different maladies) which were collected in a subsequent part of the study, after subjects were given price, ingredients, package claims, and color pictures of the products. These previous analyses have identified a strong "advertising-market share" component separating brands with higher market share and higher advertising rates from lower market share, less known brands. In addition, a second aspirin vs. non-aspirin substitute component was found in such three-way analyses.
GENFOLD2 for Analysis of Preference/Dominance Data
171
4.1.2 GENFOLD2 Analysis GENFOLD2 analyses were performed for a variety of models. Some twelve subjects were dropped from the analysis due to missing data. The data were treated as interval scale and a grand mean was removed in treating the data as unconditional. We chose not to reparameterize product coordinates as a function of price, ingredients, features, etc. here since subjects were not given such information prior to eliciting their preference ratings. We also chose not to reparameterize subject ideal points as a function of background variables since previous analyses (DeSarbo, Green, and Carroll 1983) had shown very weak relationships between such preferences and collected demographic and psychographic background variables. The simple unfolding model appeared to render the most parsimonious solution in one dimension accounting for 89.4 percent of the variance. Note, a second dimension added only 3.7% additional variance and was not interpretable. Figure 1 presents the one dimensional solution estimated using GENFOLD2 with the weighting function in expression (9) with p --- 2 (p --- 0, 1, 2, 3 was tested). Figure 1 presents the "non-degenerate" one dimensional solution where both stimuli (brands) and ideal points (subjects represented by large dots " o " ) are well intermixed. Brands that load highly positive on this dimension are heavily promoted and have larger market shares. These are brands such as Anacin, Bayer, Tylenol, Excedrin, and Bufferin, whose market share percentages were 8.6 percent or greater during the time this study was taken (Crain 1982). These brands are also ones that spend more on advertising and promotion, and were better known to the subjects in the study. The brands on the negative end of the scale are lower market share items that are less well known such as Cope, Ascriptin, Vanquish, Hudson, and Datril. Hudson Aspirin tends to be a local brand, not having as much nationwide distribution as the others. Ascriptin is a speciality brand positioned to off-set the stomach upsetting nature of many aspirin based products. Similar type ingredients are also found in Cope and Vanquish which are also not advertised all that heavily. It is interesting to note that Datril, while located in the negative segment of the dimension, is somewhat separated from Hudson, Ascriptin, Vanquish, and Cope. Datril was starting to advertise heavily during this period as a cheaper alternative to Tylenol, and has since captured a rather large portion of Tylenol's market share. For the most part, the ideal points are concentrated in the positive segment of the dimension, especially heavily concentrated around Tylenol (which at the time had the largest market share of 46.7 percent). Anacin, Bayer, Excedrin and Bufferin also have several ideal points near them. Two subjects have ideal points located nearest Datril; no ideal points are found below Datril. These results were also verified by examining means and standard deviations of the preference scores.
W.S. DeSarbo and V.R. Rao
172
I
Iii ANACIN BAYER
I{ TYLENOL e
! BUFFERIN 0 O I.
DATRIL HUDSON
ASCRIPTIN
VANQUISH
COPE
Figure 1. GENFOLD2 Pain Reliever Joint Space
4.2 Residential Communication Devices
4.2.1 Study Description A massive marketing research project was recently sponsored by a major communications firm aimed at understanding preferences for various types of residential communication devices and their features, as well as finding potential markets for new offerings. This study, conducted in July of 1981, involved a total of 499 personal interviews at four geographically dispersed Consumer Opinion Forums (central location testing facilities) at high traffic shopping malls in Charlotte (North Carolina), Chicago (Illinois), Los Angeles (California), and Ocean City (New Jersey).
GENFOLD2 for Analysisof Preference/DominanceData
173
Respondents were screened in the malls on the basis of their SocioEconomic Status and Life Cycle (SES/LC). The SES/LC score consists of information concerning respondents' family income, size of family, head of household's age, education, and occupation. Eligible respondents were escorted to a completely enclosed interviewing area where 12 actual products/brands were exhibited. Respondents were read the description of each brand by the interviewer, who then demonstrated how each brand was to be used. Respondents were asked how likely they would be to buy each of the twelve devices, within the next six months, assuming the other devices weren't available. This response was measured on an eleven point scale -- 0% (Definitely Would Not Buy) to 100% (Definitely Would Buy) in increments of 10%. Respondents were then questioned about their current household telephone inventory, demographics, and dwelling information. Because of the proprietary nature of the research, the twelve brands will be identified by the letters from A-L. In addition, the product class will not be revealed. 4.2.2 GENFOLD2 Analysis In order to address several of the sponsoring firm's research objectives, a GENFOLD2 analysis of intention to buy (preference) for the twelve brands was conducted. The GENFOLD2 option of reparameterizing X as a function of brand attributes and features (X--- 130) and Y as a function of demographic variables (Y ffi Aot) was employed. In this process, information on the following technical aspects are available from the analysis: 1. a joint space of ideal points and brands displaying the heterogeneity of subjects' preferences for the different brands; 2. a common dimensionality in which to interpret how both ideal points and stimuli are interrelated; 3. an estimate of 0 which would aid in the interpretation of dimensions and brand locations; and, 4. an estimate of a which would also aid in dimensional interpretation and also provide information as to what type of subjects buy which types of brands. GENFOLD2 provides a product positioning map of both ideal points and stimuli in a common dimensionality where the constraint coefficients (0 and a ) could aid in the decision-making for optimal product(s) selection, new product development, and market segmentation. In order to present a detailed description of the results, the analysis was restricted to a subsample of fifty subjects selected randomly from the larger sample of 499. The scale of the intention-to-buy data for the twelve brands was reversed (since GENFOLD2 is a distance model) and, assuming interval scale data, each subject's adjusted intention-to-buy data were standardized to
174
w.s. DeSarbo and V.R. Rao
have zero mean and unit variance (thus the data was treated as row conditonal). Since the primary objective was to examine feature importance and market segmentation, all analyses (internal) were performed with constraints on both X and Y. The variables comprising the background information for subjects (A) were: 1.
Sex (1 --- Male, 0 = Female),
2.
Age,
3.
Marital Status (1 = married, 0 -- other),
4.
Children Living at Home? (1 ----Yes, 0 --- No),
5.
Education Score (see Table 1),
6.
Occupation Score (see Table 1),
7.
Income Score (see Table 1),
8.
Number of Phones in Home,
9.
Home Ownership Status (1 = own, 0 = rent),
10.
Dwelling (1 --- Multi, 0 = Single), and
11.
Location of Survey (0 = East (Ocean City and Charlotte), 1 = West (Chicago and Los Angeles)).
As noted, Table 1 provides the variable codings provided by the firm for Income, Occupation, and Education. The variables comprising the descriptor variables for the twelve brands (B) were: 1.
Calling Feature (1 -- send and receive calls, 0 ~ receive calls only),
2.
Repertory Dialing (1 = Yes, 0 = No),
3.
Range (in feet),
4. 5.
Speakerphone Capability (1 = Yes, 0 = No), Price,
6.
Physical Setup (1 = stand-up set, 0 -- lie-down set),
7.
Style (I -- military (resembles a walkie-talkie), 0 = cradle).
Both sets of variables were selected after consultation with the sponsoring firm. The seven variables used in B were basically all the features where measurements were readily available. Such psychophysical variables as color, smoothness, weight, etc. were not available for use, but could conceivably carry a large importance in determining preference or intention to buy. Also, while other marketing mix variables besides price could have been utilized in B (e.g., advertising expenditures), such information was not available.
175
GENFOLD2 for Analysisof Preference/DominanceData
TABLE 1
Demographic Variable Codings
EDUCATION
SCORE
OCCUPATION
SCORE
Some Grade School Grade School Graduate Some High School High School Graduate Some College College Graduate Some Graduate School Master or Ph.D.
2 8 18 37 62 80 89 96
Retired Homemaker Unemployed Laborers Operators Service Craftsman, Foreman Sales or Clerical Student Managers, Officials, or Proprietors Professionals
10 21 23 26 32 39 47 58 64
FAMILY INCOME $3,000 or Less $3,001-$5,000 $5,001-$7,500 $7,501-$10,000 $10,001-$12,500 $12,501-$15,000 $15,001-$17,500 $17,501-$20,000 $20,001-$25,000 $25,0001-$30,000 $30,001-$40,000 Over $40,000
73 91
SCORE 3 10 17 27 38 49 59 68 79 87 93 98
The analysis was performed in 1, 2, and 3 dimensions for a variety of different models with p = 0, 1, 2, 3 for the weighting scheme defined in expression (9). Degenerate solutions disappeared with p >/ 2. (This phenomena was also observed with other data sets investigated.) Based upon diagnostics such as: correlations between dimensions, goodness-of-fit, degrees of freedom, and residual plots, the two-dimensional, simple unfolding model (X and Y constrained solution) was deemed to be the most parsimonious.
W.S. DeSarbo and V.R. Rao
176
2.0
1.5 •
•
• 00
t.0 o5
~
:
•
D C - Od
•
0.5
G
• o
L •
o 0
•
•
0.0
He B -0.5 -1.0
t -0.5
A F 0.0
t 0.5
J 1.0
t.5
Figure 2. GENFOLD2 Communication Device Two-DimensionalJoint Space
Convergence at a tolerance of .0001 occurred at major iteration 14, where the weighted variance accounted for was 86.79%. Figure 2 presents the derived two-dimensional joint space of ideal points (¥) and products
(X). While the products (labelled A-L) span all four quadrants of the space, the ideal points (appearing as large dots " o " ) appear to be somewhat concentrated in the first quadrant. Products A and B, located in the third quadrant, appear to be least desirable of all products, since they are the most remote items from the nebula of ideal points. This preference structure was in fact verified by returning to the original 50x12 data A and examining column means and standard deviations where these two products appear to be least desirable. Product H also appears to be " r e m o t e " or far away from the major concentration of ideal points, although there seem to be a few subjects for whom H is their first choice. The remaining products appear to be well dispersed amidst the concentration of ideal points -- a rare nondegenerate solution in two-way unfolding.
177
GENFOLD2 for Analysisof Preference/DominanceData
0.:5
0.2 7 0.1 3
0.0
2
--
4
--0.1
--
-0.2 -0.6
6
I -0.4
I
-0.2
1
I
1
0.0
0.2
0.4
0.6
Figure 3. NormalizedProduct DescriptorCoefficients The ideal points (Y) appear to be fairly concentrated in quadrant I. Some seven subjects, denoted in Figure 2 as " o " , possess ai < 0, and are therefore characterized by "anti-ideal" points whose locations define combinations of "least" desirable stimuli. For them, preference is higher for products farther away. In order to interpret the two dimensions extracted in the analysis, one needs to examine the constraint coefficients a and 0. Figure 3 presents a plot of the variance adjusted or normalized 0 coefficients for the products. To have a highly positive coordinate on dimension one, products appear to have a high price, be military styled, and have send and receive capabilities. These features, in fact, do separate (A,B,H) from (C,D,E,F,G,I,J,K,L) from a cursory examination of the matrix B. To have a highly positive coordinate on dimension two, products appear to have send and receive capabilities, high range, repertory dialing features, a speakerphone option, and a stand-up setup. Again, from an inspection of B, this is also an accurate portrayal with the possible exception of product H, which has a very low price.
W.S. DeSarbo and V.R. Rao
178
0.3 9 3
0.2
8 0.1 6 11
0.0
2 5
-0.1 -
-0.2 -
tO
4 -0,3
I
-0.4
-0.3
s
-0.2
I
-0.1
s
0.0
0.1
Figure 4. Normalized Subject Descriptor Coefficients
Figure 4 depicts the variance-normalized constraint coefficients a for ideal points. From this plot, one can iaterpret the two dimensions from a demographic perspective. Subjects who score highly positive on dimension one tend to own their house, be married, male, own many phones, have no children, and live in a single dwelling home. Those subjects who score highly positive on dimension two are upscale education, older, own multidwelling homes, and have lower occupation status. Thus, from the theory of componential segmentation (Green and DeSarbo 1979), we have a methodology which can examine the interaction of both subjects and products. For any product(s) in the space, one could thus examine/predict the characteristics of the subjects for whom a product would be most desirable. Similarly, for any subject(s) in the space, one could examine/predict which features of which product(s) would be most desirable. These aspects are developed in the next section of the paper.
GENFOLD2 for Analysisof Preference/DominanceData
179
5. Discussion 5.1 Managerial Implications Product managers can use the results of GENFOLD2 (in the residence communications study) to explore a number of issues in product positioning and repositioning. 5.1.1 New Product Prediction Management for the sponsoring firm wished to examine GENFOLD2 predictions for five new products that were devised sometime after the original study was conducted. These five new products were coded using their corresponding features (described earlier), and predicted locations were derived. These appear as numbers 1-5 in the revised joint space in Figure 5. From the plot, one can see that new products 3 and 5 are predicted to be most desirable. A further analysis was performed where predicted Aij were obtained for these fifty subjects for these five products and first choices (most preferred) were calculated for each of the five products. The results were: Product
No. of Subjects as First ChOice
1 2 3 4 5
7 1 27 3 12.
Predicted mean scores for the five products also clearly indicated that products three and five would be the most popular. The sponsoring firm is in the process of product testing these five products. 5.1.2 Policy Simulations and Repositioning Strategies We previously noted that brands A and B were the most " r e m o t e " and least preferred brands of the twelve in the study. Let us now assume that we are the product manager for brand A, and that after seeing Figure 2, wish to consider various changes in product features in order to reposition brand A. Let us assume that this hypothetical brand manager wishes to consider the repositioning effects of three sets of changes in product A : (a). Provide send and receive capabilities for an additional $40 in price; (b). Increase the range by an additional 200' for an additional $20 in price;
180
W.S. DeSarbo and V.R. Rao
2.0
•
•
oo $
t0-
eoO • @I L
~ o •
DCe • E ~
•
BOo •
+ sj. •
0.5
G °~Oo 0
•
•
•
I
0.0H• B
-0.5 -1.0
I
-0.5
4
A I
0.0
2 ,
0.5
I
1.0
l ....
t.5
2.0
Figure 5. Mapping of Five New Products
and, (c). Add repertory dialing and redesign to military style, all at no additional cost. The effect of these changes is displayed in Figure 6. Here, line segments connect brand A in the space to its predicted repositioned location after such changes. An " A I " is used to designate the predicted location after change (a) above; "A2" for change (b); and, "A3" for change (c). It appears that the first change would reposition A the best (closest to the cluster of ideal points) of the three. Thus, GENFOLD2 allows one to perform such "what if" types of policy simulations to examine spatially the respective repositioned locations. 5.1.3 Target Positioning/Repositioning The sponsoring firm also had interest in the design of an optimal product or products. Since the ideal points in the original analysis appear to be fairly concentrated, let us assume they comprise one market segment. The model states that the closer a product or stimulus is to an ideal point(s), the
GENFOLD2 for Analysis of Preference/Dominance Data
181
2.0~
I+5 •
• Oo
1.0
•
"Ool~ i0
II
~.. oj"
0.5
G
0;
o K 0
•
• 0.0-
-0.5 -1.0
°L •
°c'. % ,',:"
o
AI
J I
-0.5
•
Ho
•
I
I
I
t
0.0
0.5
t.0
t.5
20
Figure 6. GENFOLD2 Policy Simulations
more preferred or the more likely that product will be purchased. Thus, one might ask the question: "What type of product would best suit this target market segment's needs?" Or, from the manufacturer's point of view, what combination of features would maximize potential or expected purchase behavior? Because of the constraifits on X and ¥, one can legitimately raise these issues of product positioning and market segmentation questions and find solutions to them. The managerial goal, then, is to explore what possible combination of features would be optimal in order to best serve this target market segment. The mean coordinate value for ideal points on these two dimensions for this target segment is (.744, .415) which we shall use as a target (since the mean has minimum sum of squares from other ideal points). That is, we wish to find what combination of features, using the X constraint coefficients #, will best produce a "hypothetical" product near H - - (.744, .415). This can be translated into the optimization problem:
W.S. DeSarbo and V.R. Rao
182
Min{ II I-IG
GO tl}
,
(38)
where G contains the optimal values (to be solved for) of the seven features used in the study. Thus, we wish to find the features, which when combined, produce a product as close to the midpoint of the target market segment. Since one has both continuous and discrete variables with a nonlinear objective function, some mixed integer nonlinear integer programming algorithm would be necessary to find the globally optimum result (e.g., Dakin's (1965) modification of Land and Doig's (1960) Branch and Bound procedure would be appropriate). However, if one treats the seven features as discrete -- only examining the various levels of each feature used in the twelve existing products -- an easier approach can be used. Assuming discrete levels of each feature, a combinatorial optimization (Lin and Kernighan 1973) routine was used to examine several locally optimum solutions of optimal bundles of features. The algorithm sequentially alters the discrete levels in G until no further improvement is obtained in the minimand in expression (38). This procedure was run for some 100 trials generating a number of locally optimum solutions. In Figure 7, the target H is represented by an " X " in the joint space at (.744, .415). The closest product generated by this combinational optimization scheme appears slightly above it as " # " . It has send and receive capability, no repertory dialing, range= 1000 feet, speakerphone option, price = 239.95, lie down setup, and cradle style. In examining the distribution of features of other "not as close" locally optimal solutions, a number of features are almost always present such as send and receive capability, range of at least 300 feet, and mid-price level. Products I and L appear to be overpriced ($299.95) for this target market segment. Aided with this information, the firm is currently examining a number of different product offerings. Needless to say that these results have to be validated with a different, and hopefully, larger sample. DeSarbo and Rao (1984) report several methods of validating these results via split sample techniques. 5.2 Future Research
We have presented a detailed description of the GENFOLD2 unfolding model and our alternating weighted least squares algorithm for fitting it. Two applications of the methodology were provided. A number of "postanalytic" analyses were conducted investigating policy simulation and optimal product positioning features associated with the GENFOLD2 methodology. More work has yet to be done. A thorough Monte Carlo analysis is currently planned to examine the behavior of the algorithm as a number of different program options and data characteristics are experimentally varied.
183
GENFOLD2 for Analysis of Preference/Dominance Data
2.0
1.5
--
•
•
O0 1.0
-
• .. ".# , e
. ".L DC 0.5-
ErJ
G
• o "
O
•
•
0
0.0H •
B A
-0.5 -I . 0
t
I
I
I
I
-0.5
0.0
0.5
4.0
1.5
2.0
Figure 7. GENFOLD2 Optimal Product Locations
While the weighted loss function does appear tentatively to provide nondegenerate solutions, obvious questions are raised as to why. What really causes degenerate solutions in unfolding? Is it a particular (and common) form of error structure found in most data sets? Does it result from a poorly determined model or fiat objective function (or loss function) response surface? Our approach seems to relieve the symptoms of the disease, but we still do not really know for certain what the disease really is. More research is needed. Another related question concerns the choice of the weighting function y,j. While some guidelines can be established to rule out certain general forms of Yij, the choice of a spec~c Yu (especially p) remains as a trial and error procedure. While applications suggest a p of 2 for ~/ij defined in expression (9), more experience with the procedure must be obtained with more data sets before this recommendation can be general. Finally, experience with more real data sets is required in order to answer many of these issues raised and to properly evaluate GENFOLD2 as a reliable methodology.
W.S. DeSarbo and V.R. Rao
184
References BENNETT, J. F. and HAYS, W. L. (1960), "Multidimensional Unfolding: Determining the Dimensionality of Ranked Preference Data," Psychometrika, 25, 27-43. BECHTEL, G. G. (1976), Multidimensional Preference Scaling, The Hague: Mouton Press. BENTLER, P. M., and WEEKS, D. G., (1978), "Restricted Multidimensional Scaling Models," Journal ofMathematicalPsychology, 17, 138-151. BLOXOM, B. (1978), "Constrained Multidimensional Scaling in N Spaces," Psychometrika, 43, 397-408. BORG, I., and LINGOES, J. C. (1980), "A Model and Algorithm for Multidimensional Scaling With External Constraints on the Distances," Psychometrika, 45, 25-38. CARROLL, J. D. (1972), "Individual Differences and Multidimensional Scaling," in Multidimensional Scaling: Theory and Applications in the Behavior Sciences, Vol. I: Theory, eds. R. N. Shepard, A. K. Romney, and S. Nerlove, New York: Seminar Press. CARROLL, J. D. (1980), "Models and Methods for Multidimensional Analysis of Preferential Choice (or Other Dominance) Data," in Similarity and Choice, eds. E. D. Lantermann and H. Feger, Vienna: Hans Huber Publishers. CARROLL, J. D. and ARABIE, P. (1980), "Multidimensional Scaling," Annual Review of Psychology, Palo Alto: Annual Reviews, Inc. CARROLL, J. D., PRUZANSKY, S., and KRUSKAL, J. B. (1980), "CANDELINC: A General Approach to Multidimensional Analysis of Many-way Arrays with Linear Constraints on Parameters," Psychometrika, 45, 3-24. CARROLL, J. D., CLARK, L. A., and DESARBO, W. S. (1984), "The Representation of Three-Way Proximity Data by Single and Multiple Tree Structure Models," Journal of Classbqcation, L 25-74. CRAIN, R. (Ed.) (1982), Advertising Age, Chicago: Crain Communications. COOMBS, C. H. (1950), "Psychological Scaling Without a Unit of Measurement," Psychological Review, 57, 148-158. COOMBS, C. H. (1964), A Theory of Data, New York: Wiley and Sons. COOPER, L. and STEINBERG, D. (1970), Introduction to Methods of Optimization, Philadelphia: W. B. Saunders. COURANT, R. (1965), D~ff'erentialand Integral Calculu~ 2nd ed., Vol. I, New York: Wiley and Sons. DAKPN, R. J. (1965), "A Tree Search Algorithm for Mixed Integer Programming Problems," Computer Journal, 8, 250-255. DAVIDON, W. C. (1959), "Variable Metric Method of Minimizations," Argonne National Laboratory Report No. ANLE -- 5990. DAVIDSON, J. A. (1972), "A Geometrical Analysis of the Unfolding Model: Nondegenerate Solutions," Psychometrika, 37, 193-216. DAVIDSON, J. A. (1973), "A Geometrical Analysis of the Unfolding Model: General Solutions," Psychometrika, 38, 305-336. DAVISON, M. L. (1976), "Fitting and Testing Carroll's Weighted Unfolding Model for Preferences," Psychometrika, 41, 233-247. DE LEEUW, J., and HEISER, W. (1978), "Multidimensional Scaling with Restrictions on the Configuration," Proceedings of the Fifth International Symposium on Multivariate Analysis, Pittsburgh. DE LEEUW, J., and HEISER, W. (1980), "Multidimensional Scaling with Restrictions on the Configuration," in Multivariate Analysis V, ed. P. R. Krishnaiah, New York: North Holland, DESARBO, W. S. and CARROLL, J. D. (1981), "Three-way Metric Unfolding.," Proceedings of the 1981 TIMS/ORSA Market Measurement Conference, Providence, R. I.: Management Science.
GENFOLD2 for Analysis of Preference/Dominance Data
185
DESARBO, W. S. (1982), "GENNCLUS: New Models for General Nonhierarchical Clustering Analysis," Psychometrika, 47, 449-475. DESARBO, W. S., CARROLL, J. D., LEHMANN, D., and O'SHAUGHNESSY, J. (1982), "Three-way Multivariate Conjoint Analysis," Marketing Science, 1, 323-350. DESARBO, W. S. and RAO, V. R. (1983), "A Constrained Unfolding Model for Product Positioning," Proceedings of the 1983 ORSA/TIMS Marketing Science Conference, Los Angeles, Cal. DESARBO, W. S. and CARROLL, J. D. (1983), "Three-way Unfolding Via Weighted Leastsquares," (Unpublished Memorandum, AT&T Bell Laboratories: Murray Hill, N.J.) DESARBO, W. S., CARROLL, J. D,, and GREEN, P. E. (1984), "An Alternating Least Squares Procedure for the Estimation of Missing Preference Data in Product Concept Testing," (Unpublished Memorandum, AT&T Bell Laboratories, Murray Hill, N.J.) DESARBO, W. S. and RAO, V. R. (1984), "A Constrained Unfolding Model for Product Positioning and Market Segmentation," (Unpublished Memorandum, AT&T Bell Laboratories: Murray Hill, N.J.) FLETCHER, R. and POWELL, M. J. D. (1963), "A Rapidly Convergent Descent Method for Minimization," Computer Journal, 6, 163-168. GILL, P. E., MURRAY, W., and WRIGHT, M. H. (1981), Practical Optimization, New York: Academic Press. GREEN, P. E. (1978), Multivariate Tools for Applied Multivariate Analysis, New York: Academic Press. GREENACRE, M. J. and BROWNE, M. W. (1982), "An Alternating Least-squares Algorithm for Multidimensional Unfolding," (Presented at the 1982 Joint Meeting of the Psychometric and Classification Societies, Montreal, Canada.) HEISER, W. J. (1981), Unfolding Analysis of Proximity Data, (Published Doctoral Dissertation, University of Leiden, The Netherlands.) HIMMELBLAU, D. M. (1972), Applied Nonlinear Programming, New York: McGraw-Hill. JOHNSTON, J. (1972), Econometric Methods, 2nd ed., New York: McGraw-Hill. KRUSKAL, J. B. (1964a), "Multidimensional Scaling by Optimizing Goodness of Fit to a Nonmetric Hypothesis," Psychometrika, 29, 1-27. KRUSKAL, J. B. (1964b), "Nonmetric Multidimensional Scaling: A Numerical Method," Psychometrika, 29, 115-129. KRUSKAL, J. B., and CARROLL, J. D. (1969), "Geometric Models and Badness-of-fit Functions," in Multivariate Analysis II, ed. P. R. Krishnaiah, New York: Academic Press. KRUSKAL, J. B., YOUNG, F.W., and SEERY, J. B. (1973), "How to Use KYST, a Very Flexible Program to Do Multidimensional Scaling and Unfolding," (Unpublished Memorandum, Murray Hill, N.J.: Bell Laboratories, 1973.) KRUSKAL, J. B., YOUNG, F. W., and SEERY, J. B. (1977), "How to Use KYST2: A Very Flexible Program to Do Multidimensional Scaling and Unfolding," (Unpublished Memorandum, Murray Hill, N.J.: Bell Laboratories.) LAND, A. H., and DOIG, A. G. (1960), "An Automatic Method of Solving Discrete Programming Problems," Econometrika, 28, 497-520. LAWSON, C. L., and HANSON, R. J. (1974), Solving Least Squares Problems, Englewood Cliffs, N.J.: Prentice-Halt, Inc., I974. LIN, S, and KERNIGHAN, B. (1973), "An Effective Heuristic Algorithm for the Traveling Salesman Problem," Operations Research, 2/,489-516. LINGOES, J. C. (1972), "A General Survey of the Guttman-Lingoes Nonmetric Program Series," in Multidimensional Scaling: Theory and Applications in the Behavioral Sciences, 11ol. 1: Theory, eds. R. N. Shepard, A. K, Romney, and S. Nerlove, New York: Seminar Press. LINGOES, J. C. (1973), The Guttman-Lingoes Nonmetric Program Series, Ann Arbor: Mathesis Press. LUCE, R. D. (1959), Individual Choice Behavior, New York: J. Wiley & Sons.
186
W.S. DeSarbo and V.R. Rao
NOMA, E., and JOHNSON, J. (1977), Constraining nonmetric multidimensional scaling coto'igurations (Tech. Rep. 60). Ann Arbor: The University of Michigan, Human Performance Center. RAMSAY, J. O. (1982), "Some Statistical Approaches to Multidimensional Scaling," Journal of the Royal Statistical Society: Series .4, 145 285-312. ROSKAM, E. E. (1973), "Fitting Ordinal Relational Data to a Hypothesized Structure," Technical Report #73MA06, Nijmegen, The Netherlands: Catholic University. ROSS, J., and CLIFF, N. (1964), "A Generalization of the Interpoint Distance Model," Psychometrika, 29, 167-176. SCHONEMANN, P. H. (1970), "On Metric Multidimensional Unfolding," Psychometrika, 35, 349-366. SCHONEMANN, P. H., and WANG, M. M. (1972), "An Individual Difference Model for the Multidimensional Analysis of Preference Data," Psychoraetrika, 37, 275-309. SPENCE, I. (1979), "A General Metric Unfolding Model," Presented at the 1979 Psychometric Society Meetings, Monterey, Cal. SRINIVASAN, V., and SHOCKER, A. D., (1973), "Linear Programming Techniques for Multidimensional Analysis of Preferences," Psychometrika, 38, 337-369. TAKANE, Y., YOUNG, F. W., and DE LEEUW, J. (1977), "Nonmetric Individual Differences Multidimensional Scaling: An Alternating Least-squares Method with Optimal Scaling Features," Psychometrika, 42, 7-67. TUCKER L. R (1960), "Intra-individual and Inter-individual Multidimensionality," in Psychological Scaling: Theory and Application~ eds. H, Gulliksen and S. Messick, New York: Wiley and Sons. YOUNG, F. W. and TORGERSON, W. S., TORSCA (1967), "A FORTRAN IV Program for Shepard-KRUSKAL Multidimensional Scaling Analysis," Behavioral Science, 12, 498. ZINNES, J. L. and GRIGGS, R. A. (1974), "Probabilistic, Multidimensional Unfolding Analysis," Psychometrika, 39, 327-350.