I am currently doing a research to measure the socio-demographic factors assciated with mental health problem (categorical).
I am more comfortable using Chi-square test to measure the factors asccociated with the mental health variable. However, I saw few papers where the researchers have used Logistic regression.
Will it be wrong, if I use Chis-quare test?
Condition of Chi-Square ---- Category Vs Metric data or Metric data Vs Metric data
Condition for multiple regression ----- Metric data Vs Metric data
Common logistic regression is used for a binomial response variable. That it, a nominal variable with two categories.
There is also multinomial logistic regression that addresses a categorical response with more than two categories. Depending on your software, this may be more difficult to conduct and interpret.
In general, it is better to put all variables into one model.
However, in your case, it should be fine to look at demographic variables individually with chi-square tests. Depending on your audience, this approach may be easier to report, interpret, and plot.
Practically speaking, one thing you want to look out for is if the demographic variables are highly correlated. For example, if most of the women are high income and most of the men are low income, for any relationship with your dependent variable, you can’t separate out the effect of gender or income. This is just something to keep an eye on, and make sure you aren’t drawing unwarranted conclusions.
Also, be sure to look at the size of the effects and their practical importance. In your case, reporting the proportion of the response categories across demographic factors will probably be the easiest to interpret for your audience. And then you have to decide if these differences are practically important. If you want, you could use use Cramer’s V as an effect size statistic for a chi-square test of association.
Chi-Square used for to identify the association between two variables.
Regression used for influence of independent variable on Dependent variable.
The answers to this question might give some insight into your question. https://www.researchgate.net/post/Does_it_happen_that_Chi-Square_and_multinomial_regression_give_different_outputs_on_the_significance_of_association_between_variables
Chi-square is a measure of association. Regression is a measure of relationship. An association cannot necessarily mean a relationship. This suggests that care must be taken when using the two statistical tools since their purposes can be different. However, if you're dealing with a linear relationship regarding two or more predictor variables as against a criterion variable, regression is most appropriate.
- Badges
- Science topic
- Similar topics
- Medicine
- Public Health