Externally conditionally studentized residual from proc mixed

[mbakare]

I am trying to use influence diagnostic tool of PROC MIXED to extract out externally conditional studentized residual. Each data point is replicated four times and I expected the leverage to be 1/4 = 0.25 but I have 0.0208. My script is as follows:

proc mixed;

         class rep gen;

         model rtwt =/ddfm=kr outp=resid residual influence(iter = 10);

          random rep gen;

          ods output Influence = studres;

run;

the output dataset "resid" includes the internally conditional studentized residual which is dissimilar from internally conditional studentized residual returned by output dataset "studres" from Influence. I expected to see the same result. Why difference in the result?

Secondly, for an outlier detection, which is much better to use between internally and externally conditional studentized residual?

I attached sample of output from output dataset "studres".

Thanks

[sld]

Based on my reading of the MIXED documentation, it does not look like you can obtain externally studentized conditional residuals. The INFLUENCE option produces internally and externally studentized marginal residuals. 

The MIXED Procedure | Details | Residuals and Influence Diagnostics 

---

the output dataset "resid" includes the internally conditional studentized residual which is dissimilar from internally conditional studentized residual returned by output dataset "studres" from Influence. I expected to see the same result. Why difference in the result?

Because they are different kinds of residuals. The output dataset "studres" from INFLUENCE contains internally studentized marginal (not conditional) residuals; those will match the internally studentized marginal residuals in the output dataset from OUTPM= , but are of course different than internally studentized conditional residuals from OUTP= .

From a practical point of view, if I couldn't obtain externally studentized conditional residuals, I would use internally studentized conditional residuals.