# How to statistically analyse the survey data of the same sample taken at two points in time

I am conducting a study on a sample of 100 people in order to test whether or not education has an effect on employment status. Basically i'm running a logistic regression where education is the independent categorical variable (i.e 1= HighSchool, 2=Bachelors, 3=Masters etc...) and the dependent variable being binary response (1= employed, 0=unemployed). I also conduct this survey again 4 years later with the exact same group of people. Hence the variable that has changed is age (meaning people in my sample are 4 years older).

How can I analyse this type of study? what are some of the tests I can conduct? Please be as detailed as possible, i do not have a strong statistics background and I would greatly appreciate any guidance you can provide :)

Thanks.

## 1 Answer

You might instead analyze change in employment status (i.e. employed$_{t=2}$ - employment$_{t=1}$) as a function of educational attainment. Your outcome then becomes ternary ($-1,0,1$), because respondents can go from employed to unemployed, have no change in employment status, or become employed. So you would likely switch from logistic regression to either a contingency table test ($\chi^{2}$ test) or a multinomial-logit (polytomous) regression model.

When using ordinal predictors such as educational attainment, I might start with some kind of generalized additive model/nonparametric regression approach to see if the relationship can be represented linearly (in my own experience using educational attainment I have found both saturation, and non-monotonic effects), or if you would be best off using some nonlinear function of educational attainment.