Using PLS path analysis, I noticed the amount variance explained in an endogenous variable (which is actually a second-order hierarchical latent construct) is 0.06 (6%) which is very small. This was accounted for by 4 predictors which serve as antecedents. Does this make any sense at all and if so, what does it suggest? Any literature to point at?
Cordially
Lot of discussion on R2 is already happened, I suggest you to go through following links
https://www.researchgate.net/post/What_are_the_measures_of_goodness_if_fit_other_than_R2_What_measures_can_be_informative_when_R2_value_is_negative
https://www.researchgate.net/post/what_is_R2_and_desirability_in_Response_surface_optimization_method
https://www.researchgate.net/post/What_is_acceptable_R2_value
https://www.researchgate.net/post/what_is_the_acceptable_R2_value
https://www.researchgate.net/post/How_should_I_handle_zero_R2_values
A very high R-square does not mean that actual predictions using the model will be accurate. For small R-square one can be certain that the model only predicts values at or near the mean with accuracy. Here is a brief article demonstrating this issue, and offering a method to improve predictive accuracy:
- https://odajournal.com/2013/09/20/maximizing-the-accuracy-of-multiple-regression-models-using-unioda-regression-away-from-the-mean/
Here is a brief conceptual review of maximum-accuracy analysis:
- https://odajournal.com/2017/04/18/what-is-optimal-data-analysis/
And, an example of this methods versus correlation/regression analysis:
- https://odajournal.com/2016/09/19/novometric-analysis-with-ordered-class-variables-the-optimal-alternative-to-linear-regression-analysis/
Thanks, Alex. In the field of marketing research, R-square becomes big deal while in some areas the focus is on whether or not a statistically significant relationship exists between variables.
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS