I am investigating if the sentiments of Facebook posts of politicians change before and after the election date. My dataset consists of around 50 politicians and their posts and I ran a sentiment analysis and transformed the dataset into a range of -365 days before and 365 days after the election date (which is at timepoint 0 on the X-Axis). To deal with outliers I aggregated the daily mean of all posts. See example below: scatterplot of sentiment Tone over time, 0 is election date

Afterwards, I wanted to see if there is a structural break at the election date and ran the Chow-test.

added two single regressions, before and after election and one overall regression

So in this case, the tone of the posts does change before and after the election – so there is a structural break. Tone F = 7.6738, p-value = 0.0005194

But what tests might be useful to follow up with? I first thought about running unpaired t-tests to see if the mean changes – but perhaps it is better to focus on the variance. Are there any guidelines that go hand in hand with the chow test? I am coming from a non-statistics background and the choice between all the different tests are giving me a hard time.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.