I reviewed a paper about a month ago, in which the authors reported the protective of vitamin E against colon cancer was about 50, which is statistically significant. I knew this could not be correct and as a result, I asked the authors to check their data. Now, the revision came back and I was told that there was nothing wrong with the data and their data analysis. I don't know what to do with it.
Any suggestions?
.
"I knew this could not be correct"
I suppose this means you know some published and well-accepted references pointing out that this alledged efficiency is either impossible or has been shown not to exist.
In this case, you should ask the authors to discuss in depth why their results are at variance with well-accepted accepted pieces of litterature.
.
Otherwise, I do not think you can reject a paper based on unpublished knowledge you might have ...
.
Peizhong -
This is not my area, I do not know what an "OR value" is, but when you say that 50 is "statistically significant," it sounds like you are looking at a p-value. Note that a p-value is a function of sample size, and does not tell you much by itself. Do you have a sample size and power analysis?
The historical development of the use of the word "significance" has been rather a mess. I suggest avoiding this by using confidence intervals instead, wherever possible.
Jim
Article Practical Interpretation of Hypothesis Tests - letter to the...
I second Fabrice and Haluk:
1) if there is trustworthy knowledge contradicting the findings, then ak for a detailed discussion
2) ask to get the raw data and try to find the reason for the high OR. It may be, for instance, that the odds (of whatever) are very low in the control/Placebo group, so that the OR can be strangely high even for relatively moderate odds in the Vit.E-Group.
Example:
Control group: 1 "good case", 7 "bad cases": odds = 1/7
Vit.E group: 7 "good cases", 1 "bad case": odds = 7/1
odds ratio: (7/1) / (1/7) = 49
p-value from chi²-test: 0.02, from fishers exact test: 0.01. (H0: OR=1)
Thank you all for the insightful suggestions, which are well taken. I will look into this manuscript today again.
- Badges
- Science topic