forum.smartpls.com

Dear SmartPLS Specialists,
1-Could you please tell me why repeating bootstrapping whit the same parameters leads to different results?Could it lead to naturally different results?(I mean for example if for the first time it indicated the significance of t values but for the next time it indicates that some t values are not significant?
2-Does necessarily increasing the samples no. lead to more accurate results?

Thanks In Advance
Yours Sincerely

Bootstrapping is a random process. It samples with replacement from the original dataset. Each sample drawn is based on a new set of random numbers. Thus, there is the chance for slight differences, especially if the number of bootstrap samples is small. Increasing this number increases the stability of the estimates. Thus, we usually recommend at least 5,000 bootstrap samples for a final assessment.

In rare occasions, two different bootstrap runs can also indicate different results in significance if the effect is on the edge of being significant. You may have a p-value in one run at 0.499 and at another run at 0.501.
However, from a substantial point of view, there is virtually no difference between the results. It is only our publication system that sometimes enforces a strict threshold of <0.05 for significance. That is also why a lot of researcher recommend reporting the exact p-values instead of *** indications that hide the precise values.

I would in both cases conclude that the effect is on the edge of being significant.