I have a number of demographic and economic variables – both categorical and continuous – for a few different groups, one is a subset of the other. Estimates derive from IPUMS microdata from which I calculated crosstabs and shares for each variable by group (for easier comparison). This is basically a comparative analysis (no complicated statistical analysis). Now I want to see if there are any notable or significant difference in the portions between groups for several of the variables or if one group is more likely than another to be, for example, male/female, ages 16-19 or have a BA degree based on shares.
To be clearer, here is an example:
I have median age and age composition data broken down into 6 categories (i.e., 16-19, 20-24, 25-44, etc.) for 3 different groups.
- Group A: North Carolina’s workforce ages 16+
- Group B: North Carolina’s workforce ages 16+ whose means of transportation to work is by car, truck, or van -- drove alone (so a portion of group A).
- Group C: workforce ages 16+ who migrated to the state
As mentioned, I calculated shares for each variable, in this example for each age cohort by group, and want to know if there is a significant or notable difference in portions between the groups. In other words, how does group A compare to group B and how does group C compare to A or B? Are people from group C, for example, more likely to be between the ages of 16 to 19 than those from group A or B?
That said, which statistical test should I use, comparison of means, comparison of portions, t-test or combination of all? How should I calculate/interpret in excel? I have SPSS (with raw data) but would prefer to do the calculations in excel since the portion data is already tabulated in table format there.
Other crosstab variables I’m looking at side by side include median age, median income, gender, educational attainment, etc. by group.
Note: I've tried to answer this question myself before posting here but am still unsure the best/proper approach for my situation. I have experience with descriptive stats with my work but less experience with association or inference. It’s been many years since my stats course in undergrad and a few years since my stats courses in grad school, which covered up to multiple regression.