Sometimes when I run a mixed model from the fit model platform and then save the residuals ans studentized residuals, a large number of rows have a value of residuals but missing value for studentized residuals and I am wondering why...
After some researches, I have found that studentized residuals can be obtainded from the residuals with this formula:
stud res [i] = res[i] / (RMSE * √(1-h[i]))
where h[i] is the i th diagonal element of the hat matrix (which can also be saved in the data table)
From what I have read the diagonal elements of the hat matrix are supposed to be less than 1.
However sometimes I observe that there are greater. Thus it is not possible to calculate the square root of negative numbers and we get missing values for studentized residuals.
Is it a particular case? Is it because I have random effects in my model? Is the calculation of the hat matrix different in that case? Should we rather focus on the residuals?