Hi All,
I am doing an analysis from a survey data, if there is significant difference in use on the internet based information on gender, age or level of education.
I have Independent categorical variable: age, gender, education, frequency and dependent categorical variable: Internet Use.
I did chi-square to look at the difference between age, gender, education across Internet use.
My question is:
1) Is there any better statistical test to see the difference in dependent categorical variable?
2) Do I need to do Multiple logistic regression with all significant variable from chisq, if I want to see the combine effect?
Thank you so much. If anyone can help me with this.
Sandeep
First, why are you treating age and education as categorical variables? Even if you asked about them in categories (ranges), the traditional approach is to convert them back into interval variables by recoding each score to the midpoint of the range, or in the case of education, to score it in terms of years of education.
Beyond that, your choice of analysis depends on how you measured the dependent variable. Can you share more about that?
David L Morgan: Thanks David. Apology for not being clear about my variables.
The Survey has variable categorized as: Age (18-40, 41-70, >79 years), gender ( M, F), and Education (high school, associate, bachelor and graduate ).
Dependent variable resource is categorized as Internet, Newspaper, Family, and healthcare provider.
How can I convert thee categorical variable into interval variable or midpoint of the range? Since data was collected as a categorical variable.
So Since Both DV and IV both are categorical, I used Chi Sq. and Multivariable Logistic Regeession. I was wondering if there is any other way to see the difference in age and gender into Internet Use?
Thanks David again for ur valuable insight.
From what you have shown here Sandeep,it seems for your study all these three are categorical variables.Y David sir is asking you this question is because we can take age as a ration ordinal nominal or categorical variable based on how you measure it in the questionnaire.
So where gender coukd be nominal the age could be ordinal as even after categorization it shows order
For the Multivariable Logistic Regression, you would have gender as categorical. age (measured at the midpoint of the category) as interval, and education (measured in years) as interval.
For the chi-sq, you could treat everything as nominal, but you would not have any statistical means for examining trends across age or education. Also, unless you have a very large sample, you would have to examine the effect of each variable separately.
- Badges
- Science topic
- Similar topics
- Statistics
- Survey