Lund's outlier test [Design Issues]

posted by Helmut Homepage – Vienna, Austria, 2008-08-07 15:37 (4636 d 15:06 ago) – Posting: # 2159
Views: 12,644

Dear Ravi!

» I don't have much knowledge about Lund's outlier test. Could you Please explain it in detail and if possible please suggest some reference.

This test was stated in a guideline issued by the FDA in 1992 (Section V. Outlier Consideration), and dropped in later editions. The original reference is:

Lund RE. Tables for An Approximate Test for Outliers in Linear Models. Technometrics. 1975; 17(4), 473–76.

» Please also mention the software on which this test can be performed with some example.

As a starter you need to calculate studentized intra-subject residuals. WinNonlin and Kinetica cannot, SAS and R can… ;-)
But I would not suggest it anyhow.
Quoting Chow and Liu (2000, pp 501–2) from their critical discussion of FDA’s guideline:

“The guidance provides a pharmacokinetic definition of subject outliers and provides possible causes for their occurrence. The guidance suggests that Lund's method (Lund, 1975) be used for oulier detection. Although Lund’s method is useful in a linear regression setting that requires statistical independence of all PK responses, this method may not be appropriate for a crossover design in which the PK responses from the same subject are correlated. Although Lund's method may be applied to the difference of the PK responses between the test and the reference formulations from the same subject in a standard two-sequence, two-period crossover design, it dose not not take into account the feature of the study design. Moreover, it does not not eliminate other nuisance effects; hence, it cannot be applied to other crossover designs.”

A remark: Chow and Liu are right in their statement of missing independence of observations; on the other hand it does not make sense to apply any outlier test on the raw T/R ratios, but only to the residual errors of the statistical model.
Useful exploratory analyses are box-plots, QQ-plots, Hotelling’s T2, or Cook’s distance.

Dif-tor heh smusma 🖖
Helmut Schütz

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

Complete thread:

 Admin contact
21,419 posts in 4,475 threads, 1,510 registered users;
online 10 (0 registered, 10 guests [including 2 identified bots]).
Forum time: Sunday 06:43 CEST (Europe/Vienna)

Nothing fails like success because you do not learn anything from it.
The only thing we ever learn from is failure.
Success only confirms our superstitions.    Kenneth E. Boulding

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