Hi there,

I have been having a lot of difficulty calculating if Cook'd values are influential. I realise that there are no exact rules, but I am trying to follow a certain procedure but I'm finding it a little confusing….

"To assess if a give D is influential of not, select a criterion using the percentile value of Fisher Ratio Distribution F(p, n-p). If the percentile is less than about 20%, the case under investigation has little apparent influence on the fitted value."

In my case;
n = 18
predictors = 3
the highest d value = .96
the regression was significant at p<.05.

Any of help on how to calculate if the percentile is less than about 20%?

Cheers