Hello experts, My experiment had 13 experimental variables and 1 dependent variable and the data were collected from 30 participants. Because the 13 experimental variables are highly correlated with each other, I use PLS to extract important factors from those variables to account for the dependent variable. This is my model: # determine the number of component ncomp=selectNcomp(plsr(data=input,trimRT~var1+var2+var3+....var13,5,validation='CV',scale=TRUE), "randomization",alpha=0.05)
# feed the number of component to function to calculate loading and coefficients plsr(data=input,trimRT~var1+var2+var3+....var13,ncomp,validation='CV',scale=TRUE) I have three question regarding plsr(): 1. Should I let the model know the dependent variable (predicted variable) collected from difference people? If so, how should I code the information of subject ID in the following function? 2. Some variables are repeated-measures and some are not. How should I code this information in the function? 3. The results from loading of predictors in one factor did not match the results of coefficients (please see attached figures). For instance, given that the variable 10 had higher loading than other variables in component 5, I expected to see the coefficient of the variable 10 was higher in terms of magnitude (regardless the sign) than other variables in component 5. Why is it not the case? <http://r.789695.n4.nabble.com/file/n4733715/1.png> <http://r.789695.n4.nabble.com/file/n4733715/2.png> Your inputs are appreciated. Thanks. Best, Veda -- View this message in context: http://r.789695.n4.nabble.com/inconsistency-between-loadings-and-coefficient-in-plsr-tp4733715.html Sent from the datatable-help mailing list archive at Nabble.com. _______________________________________________ datatable-help mailing list [email protected] https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/datatable-help
