I'm doing a research and i find myself with an independent variable of ranked data (of six groups) on an ordinal scale and a dependent variable of data in the ratio/interval scale. Kindly advice on the appropriate statistical concepts/tool/method(s) of analysing this type of data pair where the objective should be to test a hypothesis. Thank you
You can proceed with logistic regression. if the dependent variable is binary, go for binary logistic. However, if your dependent variable has more than two categories go for multinomial logistic regression. If it is ordered go for the ordinal one.
Hello, can you describe more precisely your dependent and independent variables? Or if it is just a general question, a quick guide here: https://stats.idre.ucla.edu/other/mult-pkg/whatstat/
If something unclear or specific, please ask.
Regards,
Cristian Popa. Thank you for your response. I am opening the link in a moment. Basically the question is about "whether graduates trained by a team of professors would be the same in all aspects to graduates trained by a team of graduate assistants/assistant lecturers". For example, assuming all intervening factors are kept constant, (or ignored/are of negligible effect) would the two groups of graduates take the same time in months to get a job. So the I.V is rank of faculty members from graduate assistant and professor being the senior most rank. The D.V is time taken to get a job. (of course, other factors would come into question but are ignored for this particular problem). Thank you.
Sayantani Chatterjee, thank you very much. The D.V has five categories and it is also ordered in terms of rank. So that makes it ordinal i suppose. So i will try multinomial logistic regression. Thank you
Hello, Nabali,
Sorry for the delay - out for a while :-) Maybe you solved your dilemma meanwhile? How?
However, as I see, your solution would be much simpler if 'time-to-get-a-job' is straight-continuous-measured variable (eg. days). If not - drop the intervals :-)
First check for a correlation (Spearman, here!) between type of tutor and time to find a job, if any, run the plain linear regression!
If you'd like to find if type of tutor (IV) has something to do with employment at all, create a dummy variable, say 'hired' (0 if not, 1 if else), and run the logistic regression.
In your case you may want to considered Cox regression (Time-to-event), either for extra-analysis. Draw the Kaplan-Meyer curves as well.
Cheers!
- Badges
- Science topic