I am carrying out research whereby I am examining a number of health characteristics across 3 specific age categories in a large dataset of older adults. I am a statistical novice...
My dataset is grouped into disadvantaged vs non disadvantaged adults
The mean age was significantly different between these two groups so I further split these two categories into 65-74 years, 75-84 and 85+ year old
I have carried out standard Chi squared tests to examine 6 different health characteristics (binary = presence or absence of disease) for disadvantaged vs non-disadvantaged older adults within each of these age categories
I now want to carry out logistic regression on these three age categories
Should disadvantaged status be my dependent variable? Do I need to control for age within each of my logistic regression models? All of my 'predictor' variables are dichotomous categorical variables
I think I should use a multinomial logistic regression for more than 2 categories.
https://statistics.laerd.com/spss-tutorials/multinomial-logistic-regression-using-spss-statistics.php
https://www.youtube.com/watch?v=1BL5cL8_Cyc&t=434s
Hello Lauren,
It's not clear to me what you're trying to accomplish by taking a near-continuous variable (age), collapsing it into ordered categories, to predict group membership. You apparently already have demonstrated that groups are not comparable by (mean) age.
If you're trying to incorporate age as a potential predictor (beyond that of group membership) for the six health outcomes you mentioned, I'd suggest keeping it as a continuous variable (single predictor). You should likely add a group x age interaction term to the model as well.
Good luck with your work.
Hi David Morse thank you so much for your help.
If i wanted to eliminate my collapsed age groups and compare the overall group of 2000 adults by disadvantaged vs non-disadvantaged status for the 6 health variables can I add age to my regression model considering that disadvantaged adults are significantly younger overall than my relatively advantaged adults.
Hello again Lauren,
Yes. Be sure to add in an interaction term (age x group membership; just construct a new variable by multiplying) to the model.
Good luck with your work.
- Badges
- Science topic
- Similar topics
- Citations