sudy Junior India, 20180407 11:22 Posting: # 18651 Views: 506 

Dear All, I have a discussion point for Tmax analysis using NP test. I have calculated the significant pvalue for Tmax on replicate data (T R T R) by splitting in T1 R1 and T2 R2. Diffecrence1= T1R1 and Diffecrence2= T2R2. I have used NP test on these differences. First is it a correct approach logically? As i assume, tmax is nothing but a one of the protocol time point, not a derived data. So i did not calculated the average of T1&T2 and R1&R2 = TR. Second, i am getting pvalue significant. How to interpret the data then? I have used total 80 subjects; out of these 60 subjects have tmax value same for both Test and reference (i.e. T1 R1 and T2 R2). As a result, differences are zero for 60 subjects. NP test was performed by using wilcoxon signed rank test, so no rank will be assigned to these 60 subjects as their diff are zero. Only 20 subjects have nonzero value and are very much vary from each other, due to this i am getting pvalue significant. From data, my 60 out of 80 subject have Tmax value of test product similar to reference. This is good actually. But from NP test, pvalue is significant. How to handle this case? Is there any guideline recommendation? Your suggestions are most welcome. Regards, Sudy Edit: Category changed; see also this post #1. [Helmut] 
mittyri Senior Russia, 20180409 19:44 @ sudy Posting: # 18657 Views: 339 

Hi Sudy, » I have calculated the significant pvalue for Tmax on replicate data (T R T R) by splitting in T1 R1 and T2 R2. Diffecrence1= T1R1 and Diffecrence2= T2R2. I have used NP test on these differences. » » First is it a correct approach logically? I didn't get the reason of splitting by replicate. I would follow one of the methods explained here by Detlew. Note that exact tests are preferred due to ties. — Kind regards, Mittyri 
sudy Junior India, 20180420 14:15 @ mittyri Posting: # 18708 Views: 122 

» I didn't get the reason of splitting by replicate. » I would follow one of the methods explained here by Detlew. Note that exact tests are preferred due to ties. Hi Mittyri, Sorry for delay in reply. I have splitted in two parts, as i already explained, Tmax of test and tmax of reference is not a generated data, it ia one of the protocol defined time points, So logically we have to compare tmax of test to reference. If i take the average of T&T and R&R, the resulted value will not give a clearity about Tmax of formulations. Second i case of ties, logically we can prove the similarity in terms of Tmax for test and Referecne. But statistically, it will give significant pvalue, which means there is a difference. This diffecenre is due to nonties values. So how to justify. Actually, we can not say there is a significant difference in tmax. Thanks Edit: Full quote removed. Please delete everything from the text of the original poster which is not necessary in understanding your answer; see also this post #5! [Helmut] 
mittyri Senior Russia, 20180421 13:58 @ sudy Posting: # 18713 Views: 78 

Hi Sudy, » If i take the average of T&T and R&R, the resulted value will not give a clearity about Tmax of formulations. OK, do you think splitting to T1 and T2 and comparing to R1 and R2 is more reasonable method? Why are you comparing T1 to R1 and not to R2? Or would you compare all 4 combinations? What if some will say 'no difference' and one will say 'there's a difference'? If you do not want to go with average since that kinda 'parametric' then switch to method 2 described in the link above » Second i case of ties, logically we can prove the similarity in terms of Tmax for test and Referecne. But statistically, it will give significant pvalue, which means there is a difference. This diffecenre is due to nonties values. So how to justify. Actually, we can not say there is a significant difference in tmax. For me logically means statistically proven. If you cannot justify the abscence of difference scientifically then it's just 'a feeling', not logic. So please explain why don't you want to check the difference with one of the methods described in peerreviewed papers from the link? — Kind regards, Mittyri 