Studentized & standardized residuals [Outliers]

posted by Helmut Homepage – Vienna, Austria, 2020-03-13 11:21 2a02:8388:6bc2:ce80:856f:1d4e:74d9:7171 – Posting: # 21267
Views: 1,084

Hi UdayK,

» […] outlier test using Phoenix WinNonlin Software (Studentized residual method) and R-Software (code) in the Bioequivalence 2way cross design.

“Studentized residual” is an ambiguous term. There are two flavors:
  1. Internally studentized (aka “standardized”), where
    \(\widehat{\sigma}^2={1 \over n-m}\sum_{j=1}^n \widehat{\varepsilon\,}_j^{\,2}\)
  2. Externally studentized, where
    \(\widehat{\sigma}_{(i)}^2={1 \over n-m-1}\sum_{\begin{smallmatrix}j = 1\\j \ne i\end{smallmatrix}}^n \widehat{\varepsilon\,}_j^{\,2}\)
#1 is easily accessible in Phoenix WinNonlin (divide the raw residual by the standard error). Assess only the first period. The value of the other period has the same value but with the opposite sign and is therefore, not informative. No idea how to get #2 In WinNonlin. Ask at Certara’s forum.
Both are easily obtained in SAS, R,…

In R for the fixed effects model (untested!):

mod  <- lm(log(PK) ~ sequence + subject%in%sequence + period + treatment,
                     data = yourdata)
res1 <- rstandard(mod) # 1. Internally studentized (standardized)
res2 <- rstudent(mod)  # 2. Externally studentized

Note that #2 is in general more restrictive than #1, i.e., it might be that you detect more outliers.
BTW, which outlier test are you thinking of? See also there and this thread at stackoverflow.

Cheers,
Helmut Schütz
[image]

The quality of responses received is directly proportional to the quality of the question asked. 🚮
Science Quotes

Complete thread:

Activity
 Admin contact
20,458 posts in 4,296 threads, 1,414 registered users;
online 7 (1 registered, 6 guests [including 3 identified bots]).
Forum time (Europe/Vienna): 20:30 CEST

Genius is that which forces
the inertia of humanity to learn.    Henri Bergson

The Bioequivalence and Bioavailability Forum is hosted by
BEBAC Ing. Helmut Schütz
HTML5