[{"data":1,"prerenderedAt":341},["ShallowReactive",2],{"content-query-ZuRy7xLql3":3},{"_path":4,"_dir":5,"_draft":6,"_partial":6,"_locale":7,"title":8,"description":9,"layout":10,"body":11,"_type":335,"_id":336,"_source":337,"_file":338,"_extension":339,"sitemap":340},"/documentation/algorithms-and-techniques/validity-and-model-fit/prediction-oriented-model-selection","validity-and-model-fit",false,"","Prediction-Oriented Model Selection (BIC) in PLS-SEM","The Bayesian information criterion (BIC) helps researchers select the best predictive PLS-SEM model from alternative model setups. Learn how BIC relates to CVPAT in SmartPLS.","algorithm-description",{"type":12,"children":13,"toc":320},"root",[14,23,29,36,41,57,63,68,73,79,100,106,113,118,124,129,135,140,146,151,157,197,203],{"type":15,"tag":16,"props":17,"children":19},"element","h1",{"id":18},"prediction-oriented-model-selection",[20],{"type":21,"value":22},"text","Prediction-oriented Model Selection",{"type":15,"tag":24,"props":25,"children":26},"p",{},[27],{"type":21,"value":28},"Prediction-oriented model selection criteria stem from information theory and were introduced into the partial least squares structural equation modeling (PLS‐SEM) context by Sharma et al. (2019, 2021). These criteria help researchers select the best predictive model from a pre-determined range of alternative model setups, allowing them to fully exploit the predictive capabilities of PLS‐SEM.",{"type":15,"tag":30,"props":31,"children":33},"h2",{"id":32},"bic-and-the-geweke-meese-criterion",[34],{"type":21,"value":35},"BIC and the Geweke-Meese Criterion",{"type":15,"tag":24,"props":37,"children":38},{},[39],{"type":21,"value":40},"Sharma et al. (2019, 2021) show, by means of a Monte Carlo simulation, that the in‐sample model selection Bayesian information criterion (BIC) and the Geweke–Meese (GM) criterion are useful substitutes for out‐of‐sample model selection criteria (e.g., RMSA and MAD). SmartPLS provides results for the BIC for model selection. SmartPLS does not include the GM criterion, which is based on the model-complexity-adjusted mean square error (MSE) from the saturated (full) model, because defining the saturated model in PLS-SEM is not always straightforward. Especially in advanced modeling situations (e.g., moderation and second-order models), expert judgment is needed to define the saturated model, since an automatically generated saturated model could produce misleading results.",{"type":15,"tag":24,"props":42,"children":43},{},[44,46,55],{"type":21,"value":45},"You can download and run an ",{"type":15,"tag":47,"props":48,"children":52},"a",{"href":49,"rel":50},"https://www.pls-sem.net/downloads/additional-useful-downloads/",[51],"nofollow",[53],{"type":21,"value":54},"Excel file",{"type":21,"value":56}," to compute the model selection criteria yourself, including the GM criterion.",{"type":15,"tag":30,"props":58,"children":60},{"id":59},"interpreting-bic-and-gm-results",[61],{"type":21,"value":62},"Interpreting BIC and GM Results",{"type":15,"tag":24,"props":64,"children":65},{},[66],{"type":21,"value":67},"Sharma et al.'s (2019, 2021) Monte Carlo simulations show that the BIC and GM are particularly suitable for model comparison tasks. These criteria tend to select the best model among a set of competing models. In addition, both criteria achieve a sound balance between theoretical consistency and high predictive power, even in the absence of a holdout sample.",{"type":15,"tag":24,"props":69,"children":70},{},[71],{"type":21,"value":72},"Researchers need to compare the BIC and GM values across alternative model setups and select the model that minimizes these values. Hence, when using SmartPLS, you would select the model alternative with the lowest BIC outcome.",{"type":15,"tag":30,"props":74,"children":76},{"id":75},"cvpat-as-a-complementary-approach",[77],{"type":21,"value":78},"CVPAT as a Complementary Approach",{"type":15,"tag":24,"props":80,"children":81},{},[82,84,90,92,98],{"type":21,"value":83},"As an extension, Liengaard et al. (2020) proposed the cross-validated predictive ability test (CVPAT) for predictive model comparison in PLS-SEM. This specific extension has not been implemented in SmartPLS so far (and will be the subject of future SmartPLS add-ons). So far, the ",{"type":15,"tag":47,"props":85,"children":87},{"href":86},"/documentation/algorithms-and-techniques/resampling-and-inference/cvpat",[88],{"type":21,"value":89},"CVPAT for the predictive model assessment",{"type":21,"value":91},", as proposed by Sharma et al. (2023), has been implemented in the ",{"type":15,"tag":47,"props":93,"children":95},{"href":94},"/documentation/algorithms-and-techniques/prediction-and-segmentation/predict",[96],{"type":21,"value":97},"PLSpredict",{"type":21,"value":99}," results report.",{"type":15,"tag":30,"props":101,"children":103},{"id":102},"frequently-asked-questions",[104],{"type":21,"value":105},"Frequently Asked Questions",{"type":15,"tag":107,"props":108,"children":110},"h3",{"id":109},"what-is-prediction-oriented-model-selection-in-pls-sem",[111],{"type":21,"value":112},"What is prediction-oriented model selection in PLS-SEM?",{"type":15,"tag":24,"props":114,"children":115},{},[116],{"type":21,"value":117},"It is a set of information-theory-based criteria — primarily the Bayesian information criterion (BIC) — that help researchers select the best predictive model from a pre-determined range of alternative model setups in PLS-SEM.",{"type":15,"tag":107,"props":119,"children":121},{"id":120},"why-does-smartpls-report-bic-but-not-the-geweke-meese-gm-criterion",[122],{"type":21,"value":123},"Why does SmartPLS report BIC but not the Geweke-Meese (GM) criterion?",{"type":15,"tag":24,"props":125,"children":126},{},[127],{"type":21,"value":128},"The GM criterion requires defining a saturated (full) model, which is not always straightforward in PLS-SEM, especially in advanced modeling situations such as moderation or second-order models. SmartPLS therefore reports the BIC, which Sharma et al. (2019, 2021) show to be a useful in-sample substitute for out-of-sample model selection criteria.",{"type":15,"tag":107,"props":130,"children":132},{"id":131},"how-do-i-choose-between-two-competing-models-using-bic",[133],{"type":21,"value":134},"How do I choose between two competing models using BIC?",{"type":15,"tag":24,"props":136,"children":137},{},[138],{"type":21,"value":139},"Compare the BIC values of the alternative model setups and select the model with the lowest BIC value.",{"type":15,"tag":107,"props":141,"children":143},{"id":142},"is-cvpat-available-in-smartpls",[144],{"type":21,"value":145},"Is CVPAT available in SmartPLS?",{"type":15,"tag":24,"props":147,"children":148},{},[149],{"type":21,"value":150},"Yes, but not in the specific form originally proposed by Liengaard et al. (2020), which has not (yet) been implemented in SmartPLS. Instead, SmartPLS implements the CVPAT for predictive model assessment as proposed by Sharma et al. (2023) within the PLSpredict results report.",{"type":15,"tag":30,"props":152,"children":154},{"id":153},"related-smartpls-methods",[155],{"type":21,"value":156},"Related SmartPLS Methods",{"type":15,"tag":158,"props":159,"children":160},"ul",{},[161,171,179,188],{"type":15,"tag":162,"props":163,"children":164},"li",{},[165],{"type":15,"tag":47,"props":166,"children":168},{"href":167},"/documentation/algorithms-and-techniques/validity-and-model-fit/model-comparison/",[169],{"type":21,"value":170},"Model comparison",{"type":15,"tag":162,"props":172,"children":173},{},[174],{"type":15,"tag":47,"props":175,"children":177},{"href":176},"/documentation/algorithms-and-techniques/prediction-and-segmentation/predict/",[178],{"type":21,"value":97},{"type":15,"tag":162,"props":180,"children":181},{},[182],{"type":15,"tag":47,"props":183,"children":185},{"href":184},"/documentation/algorithms-and-techniques/resampling-and-inference/cvpat/",[186],{"type":21,"value":187},"Cross-validated predictive ability test (CVPAT)",{"type":15,"tag":162,"props":189,"children":190},{},[191],{"type":15,"tag":47,"props":192,"children":194},{"href":193},"/documentation/algorithms-and-techniques/validity-and-model-fit/model-fit/",[195],{"type":21,"value":196},"Model fit",{"type":15,"tag":30,"props":198,"children":200},{"id":199},"references",[201],{"type":21,"value":202},"References",{"type":15,"tag":158,"props":204,"children":205},{},[206,235,261,287,311],{"type":15,"tag":162,"props":207,"children":208},{},[209,211,218,220,226,228,233],{"type":21,"value":210},"Liengaard, B. D., Sharma, P. N., Hult, G. T. M., Jensen, M. B., Sarstedt, M., Hair, J. F., & Ringle, C. M. (2021). ",{"type":15,"tag":47,"props":212,"children":215},{"href":213,"rel":214},"https://onlinelibrary.wiley.com/doi/full/10.1111/deci.12445",[51],[216],{"type":21,"value":217},"Prediction: Coveted, yet forsaken? Introducing a cross-validated predictive ability test in partial least squares path modeling.",{"type":21,"value":219}," ",{"type":15,"tag":221,"props":222,"children":223},"em",{},[224],{"type":21,"value":225},"Decision Sciences",{"type":21,"value":227},", ",{"type":15,"tag":221,"props":229,"children":230},{},[231],{"type":21,"value":232},"52",{"type":21,"value":234},"(2), 362–392.",{"type":15,"tag":162,"props":236,"children":237},{},[238,240,247,248,253,254,259],{"type":21,"value":239},"Sharma, P. N., Liengaard, B. D., Hair, J. F., Sarstedt, M., & Ringle, C. M. (2023). ",{"type":15,"tag":47,"props":241,"children":244},{"href":242,"rel":243},"https://www.emerald.com/insight/content/doi/10.1108/EJM-08-2020-0636/",[51],[245],{"type":21,"value":246},"Predictive model assessment and selection in composite-based modeling using PLS-SEM: Extensions and guidelines for using CVPAT.",{"type":21,"value":219},{"type":15,"tag":221,"props":249,"children":250},{},[251],{"type":21,"value":252},"European Journal of Marketing",{"type":21,"value":227},{"type":15,"tag":221,"props":255,"children":256},{},[257],{"type":21,"value":258},"57",{"type":21,"value":260},"(6), 1662–1677.",{"type":15,"tag":162,"props":262,"children":263},{},[264,266,273,274,279,280,285],{"type":21,"value":265},"Sharma, P. N., Sarstedt, M., Shmueli, G., Kim, K. H., & Thiele, K. O. (2019). ",{"type":15,"tag":47,"props":267,"children":270},{"href":268,"rel":269},"https://aisel.aisnet.org/jais/vol20/iss4/4/",[51],[271],{"type":21,"value":272},"PLS-based model selection: The role of alternative explanations in information systems research.",{"type":21,"value":219},{"type":15,"tag":221,"props":275,"children":276},{},[277],{"type":21,"value":278},"Journal of the Association for Information Systems",{"type":21,"value":227},{"type":15,"tag":221,"props":281,"children":282},{},[283],{"type":21,"value":284},"20",{"type":21,"value":286},"(4), 346–397.",{"type":15,"tag":162,"props":288,"children":289},{},[290,292,299,300,304,305,309],{"type":21,"value":291},"Sharma, P. N., Shmueli, G., Sarstedt, M., Danks, N., & Ray, S. (2021). ",{"type":15,"tag":47,"props":293,"children":296},{"href":294,"rel":295},"https://onlinelibrary.wiley.com/doi/abs/10.1111/deci.12329",[51],[297],{"type":21,"value":298},"Prediction-oriented model selection in partial least squares path modeling.",{"type":21,"value":219},{"type":15,"tag":221,"props":301,"children":302},{},[303],{"type":21,"value":225},{"type":21,"value":227},{"type":15,"tag":221,"props":306,"children":307},{},[308],{"type":21,"value":232},{"type":21,"value":310},"(3), 567–607.",{"type":15,"tag":162,"props":312,"children":313},{},[314],{"type":15,"tag":47,"props":315,"children":317},{"href":316},"/documentation",[318],{"type":21,"value":319},"More literature ...",{"title":7,"searchDepth":321,"depth":321,"links":322},2,[323,324,325,326,333,334],{"id":32,"depth":321,"text":35},{"id":59,"depth":321,"text":62},{"id":75,"depth":321,"text":78},{"id":102,"depth":321,"text":105,"children":327},[328,330,331,332],{"id":109,"depth":329,"text":112},3,{"id":120,"depth":329,"text":123},{"id":131,"depth":329,"text":134},{"id":142,"depth":329,"text":145},{"id":153,"depth":321,"text":156},{"id":199,"depth":321,"text":202},"markdown","content:documentation:algorithms-and-techniques:validity-and-model-fit:prediction-oriented-model-selection:index.md","content","documentation/algorithms-and-techniques/validity-and-model-fit/prediction-oriented-model-selection/index.md","md",{"loc":4},1784805353352]