2
$\begingroup$

I'm performing a linear regression model with respect to voter participation for one of my class papers and I was curious if I could apply the same model to various time periods and then analyze/compare and contrast the differences in the significance of the model overall and the individual variables.

For instance, let's say my voter participation model for 2020 is based off of 2019 poverty, 2019 unemployment data, 2019 home ownership rates, etc.

Could I apply the same model to 2018, 2016, 2014 based off of their respective years? I know that a time series analysis would be the ideal situation here, but I am not familiar with those techniques and additionally, some of my data are 5-year estimates and I am unsure if using those in a time series analysis is a good idea.

The comparison would look like this basically (this is a hypothetical): In the 2020 regression model the percentage of the government GPD was insignificant whilst in the 2018 model, it was significant at a 0.05 level. This could perhaps be because the size of the local government in determining voter turnout through a rational choice voter model is overshadowed by the significance of a presidential election.

Any assistance is appreciated. Thanks! :)

Improve this question
$\endgroup$

1 Answer 1

Reset to default
1
$\begingroup$

Well, for the hypothetical example text the answer is "definitely no", the difference between significant and insignificant is not really interpretable (full story here: http://www.stat.columbia.edu/~gelman/research/published/signif4.pdf)

On the other hand, there's nothing wrong with the question, sounds interesting! Probably one reasonable answer is to build a multilevel model? Model the whole data set all together, in such a way that the model is aware of the grouping by year. Something like this tadpole example: https://bookdown.org/content/3890/multilevel-models.html

There are probably other reasonable ways too! But comparing significant / not significant in fits to multiple subgroups is not one unfortunately

Improve this answer
$\endgroup$
1
  • $\begingroup$ Hi Steve, Thanks so much for the information. The first thing you linked is very helpful in understanding why I can and cannot compare the significant variables in differing models. I'll be sure to spend some time looking at that multilevel model link as well when I get the chance! $\endgroup$
    – Christian Martinez
    Commented Apr 15, 2022 at 17:35

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.