Most probabilistic based samples comparing sexual minority population to heterosexual population on health outcomes or risky behavior have statistically significantly different age distributions. For example SM aged 19-29 years comprise 50% of the sample and for heterosexuals it is 30% of the sample. When evaluating the prevalence of cardiovascular disease (CVD), for example, which is strongly associated with age, the findings may or may not show an increased risk for SM. Multivariate logistic regression is used where the dependent variable is SM or heterosexual; independent variables include age, CVD, and other factors. So the model 'controls' for age but is that enough? No concerns on the findings given the age distribution? Clearly, a test just comparing the prevalence between the two group doesn't seem adequate, right?
It depends. If you use age as age in years then I see no problem, But I don't understand why SM is the DV. Why isn't CV(has, doesn't have) the DV, if you want to know predictors for who has the disease?. Why use SM since this group comprises both people with CV disease, without CV disease and many other things. If you are interested in who has CV disease then that should be the DV not an IV? Do I misunderstand something? BTW, multivariate means more than one DV, multivariable means more than one IV. This seems to be confused all the time. See the book at this link for a discussion:
https://www.amazon.com/Lawrence-L.-Kupper/e/B003W3L2CW/ref=sr_ntt_srch_lnk_1?qid=1526619893&sr=1-1
Best, David Booth
Thanks. No you are right SM would be DV. But it would be a problem with simple test of difference between two prevalence, right.
Yep, multivariate and multivariable (those pesky could of letters).
I'm sorry but I still don't understand about the DV. Best David Booth
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Statistical Modeling