Calculations for t-test for R-square, F-square

[rohit315]

I was conducting PLS-SEM analysis in SmartPLS and came across an interesting scenario

For every R square and F square value, the software provides t-test values and corresponding p-values. Am assuming the R-square or F-square p-values are being obtained by comparing the two groups of constant only model and variables added one. Is the understanding correct? IF not, how is the t-test value and p-value obtained for the two scores?

Also, should the r-square and f-square values be interpreted based on p-values or without it?

Thanks in advance!

[jmbecker]

As for any parameter in SmartPLS the significance (t-values, p-values, confidence intervals) are obtained by bootstrapping, that is, by re-running the model several times by sampling from the original dataset. On each subsample the parameter (e.g., path coefficients, R², f², etc.) are calculated and from that standard errors and confidence intervals are constructed, rebuilding the sampling distribution of the parameter.