(Total Views: 944)
Posted On: 07/12/2020 1:18:07 PM
Post# of 154090
Re: Rex Eupseiphos #42406
Rex Eupseiphos,
Thanks for your comments. “Getting into the weeds” is a good thing as we are all working in truth discovery which will help us to make the best decisions on our investment.
So, I take it that your reference to “large” p-numbers was in the sense that they “should” have been lower than they are ??. If this is due to the fact that Dr. Patterson et-al used Dunn Kruskal-Wallis, for p-value quantification and not other method, to characterize the results is like saying: a game score of 2-0 is not enough because the game was basketball and not soccer.
So, let’s agree in one thing: the results obtained by Dr. Patterson, by the method he happened to choose to analyze his measurements where good.
In regards to the appropriateness of using the Dunn correction (one of the least powerful comparison) or any other (Tukey Dunnet Scheffe Bonferoni) to the Kruskal-Wallis test or any other test, or correction, I leave it to Dr. Patterson and the other authors of the paper. It is an interesting subject but doesn’t belong here.
And, yes 10 patients (or less) don’t provide enough power to draw a more deterministic conclusion (but, by no means, meaningless) and that is why the trials are being conducted.
In regards to data snooping: I believe that Dr. Patterson reported the main parameters of interest (at least as far as I am concerned). I would have liked he reported as well CRP and have pointed out this. He reported on lymphocytes and, for Pete’s sake, did single-cell transcriptome profiles. So I don’t think there is an issue with this.
The real question here is: Will these markers reflect in statistical outcomes?
We will get the answer really soon.
Thanks for your comments. “Getting into the weeds” is a good thing as we are all working in truth discovery which will help us to make the best decisions on our investment.
So, I take it that your reference to “large” p-numbers was in the sense that they “should” have been lower than they are ??. If this is due to the fact that Dr. Patterson et-al used Dunn Kruskal-Wallis, for p-value quantification and not other method, to characterize the results is like saying: a game score of 2-0 is not enough because the game was basketball and not soccer.
So, let’s agree in one thing: the results obtained by Dr. Patterson, by the method he happened to choose to analyze his measurements where good.
In regards to the appropriateness of using the Dunn correction (one of the least powerful comparison) or any other (Tukey Dunnet Scheffe Bonferoni) to the Kruskal-Wallis test or any other test, or correction, I leave it to Dr. Patterson and the other authors of the paper. It is an interesting subject but doesn’t belong here.
And, yes 10 patients (or less) don’t provide enough power to draw a more deterministic conclusion (but, by no means, meaningless) and that is why the trials are being conducted.
In regards to data snooping: I believe that Dr. Patterson reported the main parameters of interest (at least as far as I am concerned). I would have liked he reported as well CRP and have pointed out this. He reported on lymphocytes and, for Pete’s sake, did single-cell transcriptome profiles. So I don’t think there is an issue with this.
The real question here is: Will these markers reflect in statistical outcomes?
We will get the answer really soon.
(5)
(0)
