AngusMcLean
★★

USA,
2015-05-16 17:11
(1989 d 04:45 ago)

Posting: # 14840
Views: 5,706

## Statistical Comparison Tmax [Nonparametrics]

Phoenix WinNonlin 6.4 provides a routine for statistical comparison of Tmax from a 2 way crossover study (Latin Square, 2 periods, 2 sequences) as a non-parametric parameter and confidence intervals are provided. I do not see intrasubject and intersubject variances in the output and I have just made a post to the Pharsight site regarding this.

For bioequivalence of Cmax and AUC pamameters intrasubject and intersubject variance is provided. So I am thinking how to obtain analogous information on Tmax parameter for a two-way cross-over study. I wonder if anyone has done this.

Angus

Edit: Category changed. [Helmut]
Helmut
★★★

Vienna, Austria,
2015-05-16 18:50
(1989 d 03:07 ago)

@ AngusMcLean
Posting: # 14841
Views: 4,888

## no para­meters in non­para­me­tric tests…

Hi Angus,

» Phoenix WinNonlin 6.4 […] I do not see intrasubject and intersubject variances…

Parameters (like the mean & variance) define parametric statistics (hence, the name). Since this test is nonparametric, variance(s) don’t exist.
Don’t get confused by PHX’s output, which gives also the median of test and reference. The PE of T-R is not the difference of these medians. The test calculates all pairwise differences. Their median is called the Hodges-Lehmann estimator; the CI is calculated according to Moses1,2 (see also this thread). You will also notice that the CI is rarely symmetrical around the PE.

If you want to dive deeper into the matter:
1. Hauschke D, Steinijans VW, Diletti E. A distribution-free procedure for the statistical analysis of bioequivalence studies. Int J Clin Pharm Ther Toxicol. 1990;28(2):72–8. PMID 2307548.
2. Hollander M, Wolfe DA. Nonparametric Statistical Methods.New York: Wiley (3rd ed 2013); p. 115–4.

Dif-tor heh smusma 🖖
Helmut Schütz

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

USA,
2015-05-16 19:57
(1989 d 01:59 ago)

@ Helmut
Posting: # 14842
Views: 4,786

## no para­meters in non­para­me­tric tests…

Thanks Helmut: I will look at that paper. What we are seeking is a measure of the variability of the Tmax paramater in the test formulation and the variability of the Tmax parameter in the reference formulation. The data comes from a cross-over design.

In the absence of a distribution. One can do a frequency plot and a percentile analysis and simply compare to see if one is less variable than the other................

Angus
Helmut
★★★

Vienna, Austria,
2015-05-17 00:33
(1988 d 21:23 ago)

@ AngusMcLean
Posting: # 14843
Views: 4,836

## no para­meters in non­para­me­tric tests…

Hi Angus,

» What we are seeking is a measure of the variability of the Tmax paramater in the test formulation and the variability of the Tmax parameter in the reference formulation.

OK.

» In the absence of a distribution.

<nitpick>

In the absence of a distribution defined by mean/variance.
Of course, tmax (in NCA!) follows a – though discrete – distribution.

</nitpick>

» One can do a frequency plot and a percentile analysis and simply compare to see if one is less variable than the other…

Given the limited sample size in BE studies a histogram doesn’t tell too much. I would rather go with box plots categorized for treatments. Sometimes easier to interpret – especially if you have more than two treatments.

Dif-tor heh smusma 🖖
Helmut Schütz

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