how do we interpret the results from a Nonparametric series regression

[Gijo Thomas]

I am analyzing a dataset (n=38, total 4 groups (3 groups of 10 each and 8 controls)) on occurance of a specific condition and its duration. I do have multiple episodes of the condition and varying duration, and its is different in each of the groups, the duration is not normlly distributed. I am interested to see whether there is any significant difference between the groups?
Is it okay to go for the npregress series analysis? How do i interpret the results from npregress series?

[Clyde Schechter]

I don't know much about -npregress- and can't answer the question as you have posed it. But if you are attempting to analyze an outcome that is a duration, I wonder why you do not use survival analysis techniques.

Another approach that would be fully non-parametric is Kruskal-Wallis analysis of variance, implemented in Stata as the -kwallis- command.

In addition, I would encourage you to rethink your question. The American Statistical Association has recommended that the concept of statistical significance be abandoned. See https://www.tandfonline.com/doi/full...5.2019.1583913 for the "executive summary" and https://www.tandfonline.com/toc/utas20/73/sup1 for all 43 supporting articles. Or https://www.nature.com/articles/d41586-019-00857-9 for the tl;dr. Rather than chasing the illusion of "significant difference" rather obtain some estimate of the magnitude of the difference(s) and some estimate of the uncertainties of those estimates. In the case of duration outcomes, hazard ratios for the groups might be appropriate, or differences in median duration, depending on the specific survival analysis model you use.

Finally, consider that with 38 observations divided among 4 groups, you are not well positioned to get precise answers unless the within-group variation in duration is very, very small. It would be surprising if any analysis of such a small data set turned up convincing results.

[Gijo Thomas]

Thank you very much Sir