Predictive model selection in partial least squares path modeling (PLS-PM)

Pratyush Nidhi Sharma; Galit Shmueli; Marko Sarstedt; Kevin H. Kim

doi:10.3990/2.336

Predictive model selection in partial least squares path modeling (PLS-PM)

Pratyush Nidhi Sharma
, Galit Shmueli
, Marko Sarstedt
, Kevin H. Kim

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › Academic › peer-review

82 Downloads (Pure)

Abstract

Predictive model selection metrics are used to select models with the highest out-of-sample predictive power among a set of models. R2 and related metrics, which are heavily used in partial least squares path modeling, are often mistaken as predictive metrics. We introduce information theoretic model selection criteria that are designed for out-of-sample prediction and which do not require creating a holdout sample. Using a Monte Carlo study, we compare the performance of frequently used model evaluation criteria and information theoretic criteria in selecting the best predictive model under various conditions of sample size, effect size, loading patterns, and data distribution.

Original language	English
Title of host publication	Proceedings of the 2nd International Symposium on Partial Least Squares Path Modeling
Subtitle of host publication	The Conference for PLS Users
Editors	Jörg Henseler, Christian Ringle, José Roldán, Gabriel Cepeda
Place of Publication	Enschede
Publisher	University of Twente
Number of pages	6
ISBN (Print)	9789036540568
DOIs	https://doi.org/10.3990/2.336
Publication status	Published - 2015
Externally published	Yes
Event	2015 PLS User Conference: 2nd International Symposium on Partial Least Squares Path Modeling - The Conference for PLS Users - Seville, Spain Duration: 16 Jun 2015 → 19 Jun 2015

Conference

Conference	2015 PLS User Conference
Country/Territory	Spain
City	Seville
Period	16/06/15 → 19/06/15

Access to Document

10.3990/2.336

ArticleFinal published version, 281 KB

Proceedings of the 2nd International Symposium on Partial Least Squares Path Modeling: The Conference for PLS Users
Henseler, J., Ringle, C., Roldán, J. & Cepeda, G., 2015, Enschede: University of Twente. 414 p.
Research output: Book/Report › Book editing › Popular

Open Access
File

744 Downloads (Pure)

Cite this

Sharma, Pratyush Nidhi ; Shmueli, Galit ; Sarstedt, Marko et al. / Predictive model selection in partial least squares path modeling (PLS-PM). Proceedings of the 2nd International Symposium on Partial Least Squares Path Modeling: The Conference for PLS Users. editor / Jörg Henseler ; Christian Ringle ; José Roldán ; Gabriel Cepeda. Enschede : University of Twente, 2015.

@inproceedings{08cd3f45a9144f94aeee40151484c930,

title = "Predictive model selection in partial least squares path modeling (PLS-PM)",

abstract = "Predictive model selection metrics are used to select models with the highest out-of-sample predictive power among a set of models. R2 and related metrics, which are heavily used in partial least squares path modeling, are often mistaken as predictive metrics. We introduce information theoretic model selection criteria that are designed for out-of-sample prediction and which do not require creating a holdout sample. Using a Monte Carlo study, we compare the performance of frequently used model evaluation criteria and information theoretic criteria in selecting the best predictive model under various conditions of sample size, effect size, loading patterns, and data distribution. ",

author = "Sharma, \{Pratyush Nidhi\} and Galit Shmueli and Marko Sarstedt and Kim, \{Kevin H.\}",

year = "2015",

doi = "10.3990/2.336",

language = "English",

isbn = "9789036540568",

editor = "J{\"o}rg Henseler and Christian Ringle and Jos{\'e} Rold{\'a}n and Gabriel Cepeda",

booktitle = "Proceedings of the 2nd International Symposium on Partial Least Squares Path Modeling",

publisher = "University of Twente",

address = "Netherlands",

note = "2015 PLS User Conference : 2nd International Symposium on Partial Least Squares Path Modeling - The Conference for PLS Users ; Conference date: 16-06-2015 Through 19-06-2015",

}

Sharma, PN, Shmueli, G, Sarstedt, M & Kim, KH 2015, Predictive model selection in partial least squares path modeling (PLS-PM). in J Henseler, C Ringle, J Roldán & G Cepeda (eds), Proceedings of the 2nd International Symposium on Partial Least Squares Path Modeling: The Conference for PLS Users. University of Twente, Enschede, 2015 PLS User Conference, Seville, Spain, 16/06/15. https://doi.org/10.3990/2.336

Predictive model selection in partial least squares path modeling (PLS-PM). / Sharma, Pratyush Nidhi; Shmueli, Galit; Sarstedt, Marko et al.
Proceedings of the 2nd International Symposium on Partial Least Squares Path Modeling: The Conference for PLS Users. ed. / Jörg Henseler; Christian Ringle; José Roldán; Gabriel Cepeda. Enschede: University of Twente, 2015.

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › Academic › peer-review

TY - GEN

T1 - Predictive model selection in partial least squares path modeling (PLS-PM)

AU - Sharma, Pratyush Nidhi

AU - Shmueli, Galit

AU - Sarstedt, Marko

AU - Kim, Kevin H.

PY - 2015

Y1 - 2015

N2 - Predictive model selection metrics are used to select models with the highest out-of-sample predictive power among a set of models. R2 and related metrics, which are heavily used in partial least squares path modeling, are often mistaken as predictive metrics. We introduce information theoretic model selection criteria that are designed for out-of-sample prediction and which do not require creating a holdout sample. Using a Monte Carlo study, we compare the performance of frequently used model evaluation criteria and information theoretic criteria in selecting the best predictive model under various conditions of sample size, effect size, loading patterns, and data distribution.

AB - Predictive model selection metrics are used to select models with the highest out-of-sample predictive power among a set of models. R2 and related metrics, which are heavily used in partial least squares path modeling, are often mistaken as predictive metrics. We introduce information theoretic model selection criteria that are designed for out-of-sample prediction and which do not require creating a holdout sample. Using a Monte Carlo study, we compare the performance of frequently used model evaluation criteria and information theoretic criteria in selecting the best predictive model under various conditions of sample size, effect size, loading patterns, and data distribution.

U2 - 10.3990/2.336

DO - 10.3990/2.336

M3 - Conference contribution

SN - 9789036540568

BT - Proceedings of the 2nd International Symposium on Partial Least Squares Path Modeling

A2 - Henseler, Jörg

A2 - Ringle, Christian

A2 - Roldán, José

A2 - Cepeda, Gabriel

PB - University of Twente

CY - Enschede

T2 - 2015 PLS User Conference

Y2 - 16 June 2015 through 19 June 2015

ER -

Predictive model selection in partial least squares path modeling (PLS-PM)

Abstract

Conference

Access to Document

Fingerprint

Research output

Proceedings of the 2nd International Symposium on Partial Least Squares Path Modeling: The Conference for PLS Users

Cite this