1

I have performed a study evaluating the diagnostic performance of a new test (T) (binary outcome) for a disease (D). Logistic regression was performed with the disease state as a dependent variable, the performance of the new test (T), and a few previously published potential risk factors were used as the independent variable. It showed only the new test (T) is a significant variable.

I then used the independent sample t-test to evaluate the risk factors. It was significant for one of the risk factors (R1). I then repeated the logistic regression without the performance of the new test (T), R1 is now a significant variable.

Can I conclude R1 is a risk factor for D but because of the remarkable performance of the new test (T), it does not contribute in diagnosing D; or it is possible that I just don't have enough sample size to demonstrate the significance of R1 in the logistic regression model?

kjetil b halvorsen
  • 77,844
George
  • 11
  • Maybe you can give some contextual information? Sample size, whar are your variables, concretely maybe show some plots ... correlation between predictors ... at present your question is too abstract, so only generalities can be said ... but there are many similar Qs here, maybe https://stats.stackexchange.com/questions/404275/inconsistency-between-multivariate-logistic-regression-and-independent-t-test/404284#404284 https://stats.stackexchange.com/questions/161278/how-to-deal-with-linear-regression-intercepts-with-high-p-values-in-dichotomic-c/161603#161603 – kjetil b halvorsen Dec 18 '22 at 13:39

0 Answers0