forum.smartpls.com

Why not thinking to add the calculation of the p-value? It will really help a lot instead of obtaining it by using the t-value

I also think this could be a good idea.

I would rather prefer confidence intervalls (bias corrected and accelerated) - the p-value assumes that your bootstrap distribution is normal (or t-type) and it is generated under Ho. In addition, the width of the CI provides you additional information.

I think it would be useful to add the p-value.

Agreed - p value should be in the output. Is there a reason for not showing the p value?

I agree with Andreas Pufall that the data might not be normal. What about adding tests of normality? Something like in GLM ANOVA in NCSS - where there is a normality test, ANOVA and Kruskal-Wallis test, so one may decide when to use Kruskal-Wallis instead of ANOVA based on the normality test. (2developers: please think twice when choosing which normality test to implement.)

But I would still appreciate another column (next to t-statistics) with p-values.

Suggestion to make everyone happy.

If the data is generally normal - pick a skewness/curtosis or let the user enter/select them, then report p-values otherwise report confidence intervals.

Jose

What about adding tests of normality?







__________________
fut 15 coins

All SmartPLS 3 bootstrap reports include p-values (and t-values and confidence intervals and bias corrected confidence intervals) from the very beginning. Just take a look.