Prediction Oriented Classification in PLS Path Modeling

Squillacciotti, Silvia

doi:10.1007/978-3-540-32827-8_10

Cited by 22 publications

(28 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…8 The latent class-detection approach was fi rst introduced to marketing scientists through the ML-SEM-based fi nite-mixture model, 10 whereby segmentation is performed by taking into account the defi ned model and the implied relationships among variables. 7 However, owing to the high dependency of ML-SEM (latent class approach) on multivariate normality and large sample sizes, several scholars prefer the PLS-SEM (latent class approach) because it is often better adapted to empirical data. 9,11,14 Over the last decade, several latent class approaches have been developed in the domain of PLS-SEM in order to deal with unobserved heterogeneity (see Sarstedt, 2008 8 ).…”

Section: Capturing Heterogeneous Consumer Behaviourmentioning

confidence: 99%

“…There exist, however, two different SEM methods for such a purpose: the maximum-likelihood approach to structural equation modelling (ML-SEM) and the partial least squares approach to structural equation modelling (PLS-SEM). 7 Whereas the model-based approach for detecting latent classes in ML-SEM has been in use since 1990s, the same approach in PLS-SEM has only recently received attention in marketing research literature. 8 Consequently, over the past decade, several techniques in the PLS-SEM domain for modelbased segmentation have arisen, one of which is the REBUS-PLS (response-based procedure for detecting unit segments in partial least squares) proposed by Esposito Vinzi et al 9 The purpose of this article is fi rst to provide a rationale for the use of model-based segmentation, and then to exhibit how a post hoc model-based segmentation can be performed with the use of REBUS-PLS on a set of consumer-behaviour data.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Model-based post hoc segmentation (with REBUS-PLS) for capturing heterogeneous consumer behaviour

Mehmetoglu¹

2011

J Target Meas Anal Mark

View full text Add to dashboard Cite

There is in the marketing literature a consensus that consumer behaviour varies across segments. Capturing this heterogeneity poses a considerable challenge for marketing scientists. Several researchers have consequently formulated a priori and post hoc techniques for detecting consumer heterogeneity. One such recent post hoc technique is model-based segmentation enabled by REBUS-PLS (response-based procedure for detecting unit segments in partial least squares), which has been developed within the framework of partial least squares approach to structural equation modelling. This article argues that model-based segmentation, as opposed to the alternative techniques, is a more objective and reliable approach for capturing heterogeneity in consumer behaviour. Moreover, this article illustrates how a model-based segmentation can be performed using REBUS-PLS through a set of consumer data, and explains conceptually the advantages of this statistical procedure. From a practical point of view, the article also proposes that model-based segmentation with REBUS-PLS or similar statistical procedures can contribute to the appeal of segmentation studies in the eyes of marketing managers.

show abstract

Section: Capturing Heterogeneous Consumer Behaviourmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Model-based post hoc segmentation (with REBUS-PLS) for capturing heterogeneous consumer behaviour

Mehmetoglu¹

2011

J Target Meas Anal Mark

View full text Add to dashboard Cite

show abstract

“…As a distance-based segmentation method, the PLS prediction-oriented segmentation (PLS-POS) method builds on earlier work on distancemeasure-based segmentation-that is, the PLS typological path modeling (PLS-TPM) approach (Squillacciotti 2005) and its enhancement, the response-based detection of respondent segments in PLS (REBUS-PLS) (Esposito Vinzi et al 2008). To extend the distance-measure-based PLS segmentation methods (including overcoming the methodological limitation of PLS-TPM and REBUS-PLS being applicable only to PLS path models with reflective measures (Esposito Vinzi et al 2008;Sarstedt 2008)), the PLS-POS algorithm introduces three novel features: (1) it uses an explicit PLS-specific objective criterion to form homogeneous groups, (2) it includes a new distance measure that is appropriate for PLS path model with both reflective and formative measures and is able to uncover unobserved heterogeneity in formative measures, and (3) it ensures continuous improvement of the objective criterion throughout the iterations of the algorithm (hill-climbing approach).…”

Section: Overviewmentioning

confidence: 99%

“…To reassign observations, PLS-POS builds on the idea of Squillacciotti (2005) and Esposito Vinzi et al (2008) to use a distance measure. We propose a new distance measure that is applicable to both reflective and formative measures and accounts for heterogeneity in the structural and the formative measurement model.…”

Section: Distance Measurementioning

confidence: 99%

“…Second, in each iteration step, the algorithm changes the assignment of only one observation and calculates the group-specific PLS estimates of all observations and their new distance measures. Thus, in contrast to the alternative distance-based PLS segmentation approaches suggested in the literature to date (e.g., Esposito Vinzi et al 2008;Squillacciotti 2005), PLS-POS avoids moving a sizeable set (more or less) of similar candidates from one group to another without improving the objective criterion. Third, owing to the implementation of a hillclimbing approach, PLS-POS could face the problem of ending in local optima.…”

Section: Algorithmmentioning

confidence: 99%

See 1 more Smart Citation

Implementation Success

Management Accounting &Amp; Control Scales Handbook

View full text Add to dashboard Cite

The Pathmox approach for PLS path modeling segmentation

Lamberti

Aluja

Sánchez³

2016

Appl Stoch Models Bus & Ind

View full text Add to dashboard Cite

Modeling has very often failed to live up to expectations, mostly because of the difficulty of comprehending relationships within phenomena and expressing them in mathematical models. Reality is frequently too complex to be reflected in a single model. This is often the case of marketing research, where variables relating to socioeconomics, psychographics or consumption constitute potential sources of heterogeneity. In such cases, the assumption of "one model fits all" is unrealistic and may lead to inaccurate decisions. Thus, heterogeneity is a major issue in modeling. Once a model has been fitted to a complete data set that fulfills all validation criteria, it is difficult to establish whether it is valid for the whole population or it is merely an average artifact from several sub-populations. The purpose of this paper is to present the Pathmox approach to deal with heterogeneity in Partial Least Squares Path Modeling (PLS-PM). The idea behind Pathmox is to build a tree of path models that have look-alike structure as a binary decision tree, with different models for each of its nodes. The split criterion consists of an F statistic comparing two structural models, which test the identity of two regression models. In order to ensure the suitability of the split criterion, a simulation study was conducted. Finally, we have applied Pathmox to a survey that measured Satisfaction among Spanish mobile phone operators. Results suggest that the Pathmox approach performs adequately in detecting PLS-PM heterogeneity. Keywords: Heterogeneity, Partial Least Squares Path Modeling, Segmentation, Model Comparison, Segmentation Trees. IntroductionIn survey research, it is very often taken for granted that a single model will represent the whole population that is, homogeneity is presumed for all the individuals. However, it is sometimes reasonable to cast doubt on such supposed homogeneity, and it is even logical to assume that groups have different behaviors. This is especially true of market research studies, where different customers/users may have different needs ([1], [2], [3] and [4]). If we do not consider the possible existence of segments in our population, the one-size-fits-all solution might be inappropriate and may generate inaccurate and/or inadequate results. This is particularly the case of Structural Equation Modeling (SEM) studies, where variables of socio-demographics, psychographics or consumption may stem from potential sources of heterogeneity. As [3] pointed out SEM has become a quasi-standard in marketing research. Its main advantages are: it allows to model latent concepts together with observed variables; researchers can test their current knowledge of phenomena using information gathered from the data; and they can validate theories, establish the main drivers of a phenomenon, measure intangibles, and advance new hypotheses. Regarding the problem of customer satisfaction, Partial Least Squares Path Modeling (PLS-PM) has become a reference methodology ([5] [13]). PLS-PM is based on an iterat...

show abstract

Prediction Oriented Classification in PLS Path Modeling

Cited by 22 publications

References 18 publications

Model-based post hoc segmentation (with REBUS-PLS) for capturing heterogeneous consumer behaviour

Model-based post hoc segmentation (with REBUS-PLS) for capturing heterogeneous consumer behaviour

Implementation Success

The Pathmox approach for PLS path modeling segmentation

Contact Info

Product

Resources

About