Dear all,
I want to assess predictive power of my model. It has 5 latent variables, and the maximum number of arrows pointing at one construct is two. I have around 120 observations in my sample. I'm not sure how many folds should I choose. It is suggested in the literature (Hair j., Risher J., Sarstedt M., Ringle C. (2019). When to use and how to report the results of PLSSEM. European Business Review, 31(1), 224.) that every fold(subset) has to meet minimum sample size guidelines. Since my sample is modest, is it enough to use, for example, only 3 folds? Or should I use more, 4 or 5, or even more... Also, the results differ depending on the number of folds I set. So I don't know which number of folds to chose.
Thank you in advance for your time and answers.
Kind regards,
Katarina
PLSpredict for assessment of the model's predictive power

 PLS Junior User
 Posts: 5
 Joined: Thu Feb 21, 2019 9:44 am
 Real name and title: PhD Katarina Njegic Assistant Professor

 SmartPLS Developer
 Posts: 1079
 Joined: Tue Mar 28, 2006 11:09 am
 Real name and title: Dr. JanMichael Becker
Re: PLSpredict for assessment of the model's predictive power
Usually it is recommended to use 10 folds. In this case your training data is 90% of your total sample size. This training data should meet the requirements for minimum sample size.
A good new source for applying PLSPredict is also the following:
Shmueli, G., Sarstedt, M., Hair, J.F., Cheah, J.H., Ting, H., Vaithilingam, S., & Ringle, C.M. (2019). Predictive model assessment in PLSSEM: Guidelines of using PLSpredict, European Journal of Marketing, 53(11), 23222347.
A good new source for applying PLSPredict is also the following:
Shmueli, G., Sarstedt, M., Hair, J.F., Cheah, J.H., Ting, H., Vaithilingam, S., & Ringle, C.M. (2019). Predictive model assessment in PLSSEM: Guidelines of using PLSpredict, European Journal of Marketing, 53(11), 23222347.
When choosing a value for k, researchers have to ensure that the training sample in a single fold still meets the model’s minimum sample size requirements (Kock and Hadaya, 2018). For example, with 200 observations and k=5, each fold’s training sample has 160 observations. Hence, the minimum sample to estimate the underlying model must be 160 (or higher). Predictive studies typically set k to 10 (Shmueli et al., 2016). We recommend following this convention as long as the minimum sample size requirements are met. If a k of 10 does not result in the minimum required training sample size, researchers should choose a higher k value to increase the size of the training sample in each crossvalidation run.
Dr. JanMichael Becker, University of Cologne, SmartPLS Developer
Researchgate: https://www.researchgate.net/profile/Jan_Michael_Becker
GoogleScholar: http://scholar.google.de/citations?user ... AAAJ&hl=de
Researchgate: https://www.researchgate.net/profile/Jan_Michael_Becker
GoogleScholar: http://scholar.google.de/citations?user ... AAAJ&hl=de

 PLS Junior User
 Posts: 5
 Joined: Thu Feb 21, 2019 9:44 am
 Real name and title: PhD Katarina Njegic Assistant Professor
Re: PLSpredict for assessment of the model's predictive power
Thank you very much for the answer, professor! It is very helpful! I understand it better now.
Best wishes,
Katarina
Best wishes,
Katarina