Nazir Ahmad Tookhy, I have some questions.
Thank you for clarifying.
Bruce Weaver
Thank you for reply, Yes it is categorical.
In brief, I am going to determine risk factors associated with rumen fluke infection. I collected faecal samples and conducted Flukefinder method to see the prevalence. I also prepared a questionnaire and asked some questions regarding farm management data and other risk factors from the farm owners.
First I did Chi square that which variables are significant and not significant. I check the Flukefinder with all risk factors such as (age, sex, farm type, animal breed, production type etc...). I attached here my Univariate analysis table. Now when I want to find the OR and 95% CI for variable with three option, I can not do it and the result says (a. Risk Estimate statistics cannot be computed. They are only computed for a 2*2 table without empty cells).
I really need to know the way that I am doing is correct or no and how to find OR and CI for variable with three or more options.
Note: After Univariate analysis, the variables which were significant, i put in binary logistic regression and here none of them were significant.
Thank you
Bruce Weaver Also my another question is if my variables are not significant in Binary logistic regression, Should I make a table and put in my thesis or ignore it.
Nazir Ahmad Tookhy, the following sections of Frank Harrell's Author Checklist address your question about variables that are not statistically significant in univariable models.
- https://discourse.datamethods.org/t/author-checklist/3407#arrow_rightuse-of-stepwise-variable-selection-24
- https://discourse.datamethods.org/t/author-checklist/3407#arrow_rightlack-of-insignificant-variables-in-the-final-model-25
- https://discourse.datamethods.org/t/author-checklist/3407#arrow_rightoverfitting-and-lack-of-model-validation-26
Mike Babyak also addresses these issues in his very readable article on overfitting:
- https://people.duke.edu/~mababyak/papers/babyakregression.pdf
HTH.
- Badges
- Science topic