Hi Friends,
I have found expression of a gene ( continuous variable) is associated with prognosis in colorectal cancer from 7 different datasets a by logistic regression. Thus, I have 7 odds ratio from different dataset. How can I calculate the 'pooled odds ratio' for this gene from these 7 individual odds ratio?
Best,
Nirmalya
In this case, the datasets should be pooled and run as a single data set. That is the only way you can pool the ORs. I am attaching a paper on a related issue that you may find helpful. Best wishes, David Booth
Hi David,
I used 7 NCBI-GEO microarray datasets which used same Affimetrix platform HUG133. Do I need any kind of normalization before use those as a single one (Fluorescence values are natural log converted in the datasets) . If I find they are normally distributed, can I run the combined dataset for Logistic regression without any normalization?
Are you suggesting this paper? Preprint Boosting and Lassoing Cancer Risk Factors: New Prostate Canc...
Because, the attached one is a letter to editor from you.Thanks,
Nirmalya
Sorry got in a hurry. Yes I am suggesting that lasso or boosting is a better approach than only logistic regression by itself, As you can see I got more information that way than Matt did with ordinary logistic regression. Lasso works when p>>n too I attached a paper on that here. I have read Knudsen's microarray stuff but I am not an expert by any means. I would try normalization both ways but my guess is it won't make a difference. Good luck and best wishes, David
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Meta-Analysis