Re: Intake of Soy Isoflavones on Breast Cancer Recurrence.

We discussed this on the web MD Diet Debate board and Heretic posted the following:

Something seems to be wrong with their stats. If you look at the table 2:



(

)


The "p for trend" in the 3 columns out of 4 is 0.46,0.87 and 0.76 and only in one column (the third) it is 0.02! That is in spite of the fact that the statistical samples in each subgroup (bin) was similar, that is about 20 cases out of 50 to 70 total.

To remind us what it means, P is the probability that the results are purely by a chance. P value of 0.05 or less is regarded as the accepted threshold of being "statistically significant".

I suspect that their p=0.02 was some calculation artefact (*) or error and all results are in fact not statistically significant.

Heretic

*) Note that the sample size of 20 cases implies about plus minus 20% error bar while the effect itself is of the order of -12% to -33% in terms of Hazard Ratio reduction - comparable to the error bar itself. That is more consistent with the p of 0.46-0.87 rather than p=0.02 Tweet