I am conducting a research where all my variables are categorical except for age. I have one dependent and 8 independent variables. I need to know which test to apply. After bit of research I have came to know that I need to run logistic regression but what else should I look for?Should I check normality of data?Should I check for independence of mean? Should I apply Chi-square test or Mann Whitney Test? I basically am a bit naive here as a newbie in research?Will appreciate any kind of help?
How many categories in your DV is important as is if they might be ordinal.i am attaching a recently accepted paper that may be helpful to you. I also added some course notes with examples. Best wishes, David Booth PS if you have more two nominal categories contact me as I have software for that.
David,thanks for response.All the categorical variable under study have two categories that are yes or no.It is a medical based research.I am sorting the relationship between status at discharge(mortality being the primary) and my independent variables.
The following can be very helpfull for you:
https://www.researchgate.net/publication/350818602_Questionnaire_Analysis_Roadmap
Regards
Given the dependent variable is binary, binary logistic regression is what you want to use. There are no parametric assumptions for the independent variables. I suggest coding the categorical independents as -1, +1 and recoding age as deviation from the mean (deviation from median if it is skewed) so that you can use multiplication to form interaction (moderator variable) terms (e.g., age-gender-interact = age*gender) if you are interested in in testing for those effects in your model. I used logistic regression many times in our work on in-hospital resuscitation.
Hello Laiba Fazal. Given a dichotomous outcome variable, I agree with Francis C. Dane's suggestion to use binary logistic regression. But given the number of explanatory variables, I am concerned that you might violate the rule of thumb that Frank Harrell calls the 20:1 rule. What is your sample size, and how many events and non-events are there?
For more info, search for 20:1 on this webpage for more info:
- https://discourse.datamethods.org/t/author-checklist/3407
HTH.
Hi, Mr. Bruce. My sample size is 385. My dependent variable is dichotomous and my independent variable are dichotomous as well as trichotomous. I am thinking of finding descriptive statistics for each variable. Check the correlation co-efficient of my dependent variable with each independent variable using chi-square test for independence. Lastly, apply regression. I can find mean for my variables that are nominal whereas, median for categorical variable.What other test can I apply?Should I check for normancy in my variables?Do you think so it is a right approach?
What is your research question? How do you expect the predictor variables to relate? Are you interested in bivariate associations? Causation? etc. The lack of information that you provided means the answers above are just people guessing about what you want.
Hi Daniel, my dependent variable is mortality and my independent varaliable are age,gender,hospital stay in days,use of steroids...etc. I want to check the correlation between mortality and all independent variable.
- Badges
- Science topic
- Similar topics
- Mathematics
- Algebra
- Matrix