I want as part of my job, research the faceturs associated with HIV in 2015 and 2017 and compare the odds ratio over these two periods. My concern is that the sample is not the same. I would like to know if there is an opportunity to do it.
I agree with Asaad Sharhani . It does not matter if the sample are not the same. Odds ratio is a is a probability value, regardless of the samples size.
Thank you for your answer. I have two survey which involved MSM people. The first has been taken place in 2015 and the second in 2017. I wan't to know if the sexual behaviors of those population are changed . To do that, i use logistics regression and i want to compare the Odds ration throught a statistic test. The population of this two survey are not the same.
okay now I understand for such reasons you can use odds however it could sound more professional if you could include the 95% confidence interval as well.
so using logistic regression you expect to have binary outcomes? if yes what is it? and what is your outcome measures?
Ideally, like should be compared with like. But your data come from two sources/ two different universe, since the samples are from two different years. As it is not a pre-test, post-test design, rather comparison of interventions in 2015 and 2017, so sample size difference is not that important. But as has been suggested by some friends, please keep the confidence level in each sample equal and conventionallay above 95% for calculating the odds first. Please control other variables that may confound.
I need to do statistical signifinace test between the results of my replication study and the original study. There are 30 survey questions with three responses to each. I have percentages of each response. How can I compare these percentages between two studies.
Attached is the data. (There are also mean and SD. Those were calculated based on the reponses on Likert scale of 1-5)