Our study is a cross-sectional study with a total sample size of 30. The dependent variable (presence or absence of MP components in the detected blood clot samples) is a binary categorical variable, with 24 samples testing positive for MP and 6 samples testing negative. The reviewers suspect that the wide 95% confidence interval of the odds ratio may indicate an exaggerated association, potentially rendering the logistic regression model unreliable. In this case, can we still use the logistic regression model for analysis? It seems that a multiple logistic regression may not be appropriate, but can we use the results of a single-factor logistic regression? Given the small frequency of negative results and an 80% occurrence rate of positive results, could we use an adjusted Poisson regression model for this binary outcome? If the adjusted Poisson regression model can be used, would it be unusual to compare the results of a single-factor logistic regression with those of the Poisson regression?
Asked
Active
Viewed 35 times
0
-
2I don't know if there is anything of consequence that can be assumed from 30 observations. – Shawn Hemelstrand Feb 25 '24 at 15:05
-
Because it is an emerging research area, the sample size is generally small. – zhiheng yi Feb 26 '24 at 08:26