Dear fellow researchers,
As a part of my study, I am calculating the odds ratio in Head and Neck cancer patients. At first, I calculated the odds ratio separately for each risk factor (smoking, alcohol, tobacco chewing, oral sex). But then I realized some patients were exposed to multiple risk factors, so in such cases how do I calculate the OR? Should I adjust the values to one risk factor? Kindly share your thoughts and suggestions.
Thanks in advance!
Hi there Pushkal! I hope you're doing well and are staying healthy. If there are many different variables, and you are considering the total impact. I'd consider going for a chi-square test of independence.
When data is investigated on a 2x2 table, the simplest way to approach their relationship as binary variables is odd’s ratio which singles out a value that represents their relationship, when we extend 2x2 to become I x J’s the proper methods of understanding the relationship of values turns to the Chi-squared test of independence, which would have (I-1)x(J-1) degrees of freedom in their representation. This tells us that the concepts of independence for tables of I & J >2 are represented through multiple layers of ways it can deviate from independence, this further means that the relationship cannot be represented further as a single value such as is present in the 2x2 tables and needs to be presented by (I-1) x (J-1) values of odds ratios, difference in proportion and relative risk.
I hope this was helpful!
This is a perfect example of where 'naive' multi-variate logistic regression is perfectly reasonable, given that there is strong evidence for each of these variables as an independent risk factor. You can code each variable as a dummy variable and then just do multivariate logistic regression to obtain adjusted odds ratios
@jack Henry. Excellent comment but many confuse multivariate with multivariable Please be careful. Best wishes, David Booth
You can perform a multivariate analysis according to the type of dependent variable evaluated.
Thank you so much for responding in such a short period of time Mauricio López-Espejo David Eugene Booth Jack Henry and Ahmad Mhanna
I will consider your suggestions and will get back if I have any more doubts.
Good day!
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics