Why does setting permI to 0, 20(default) or 100 give the same result and what does it mean?

In all three cases c(0,20, 100) , the summary statistics are the same

oplsda <- opls(df, as.factor(df$y), predI = 1, orthoI = NA, permI = 0)

OPLS-DA 346 samples x 9144 variables and 1 response standard scaling of predictors and response(s) R2X(cum) R2Y(cum) Q2(cum) RMSEE pre ort Total 0.184 0.944 0.79 0.119 1 4

Hi, As explained in the vignette (sections 4.3 and 4.5.1), the permI argument is used to evaluate the significance of the model (i.e. to avoid using overfitted models): more precisely, permI models are computed after random permutation of the Y response and the proportion of the corresponding Q2Y values above the Q2Y from the initial model (without permutation) are computed and indicated in the summary statistics (which are not the same) as the pQ2 value. This proportion is expected to be low for a valid (significant) model. For example, if permI is set to 20 (default) and all Q2Y values of the models with randomized responses are below the Q2Y from the initial model, pQ2 = (1 + 0) / 20 = 0.05. It is very important that permI is set to at least 20 and that the resulting pQ2 is below 0.05 for an (O)PLS(-DA) model to be used. Best, Etienne.