1
$\begingroup$

I am trying to build an economic model using multiple regression, and I am not sure how to remedy seasonal effects. I am collecting data across several different variables, and building three models from the same data: annually, quarterly, and monthly. Intuition would tell me the monthly or quarterly model would be the best, but so far the annual model looks much better, perhaps because of a large seasonality effect. If I add dummy variables in the quarterly and monthly models, will that solve my seasonality problem?

$\endgroup$
2
  • $\begingroup$ Difficult to comment without any sight of the data or the results. It's entirely plausible that annual data work better if your unstated predictors don't capture the seasonality. Indicator variables for periods of the year will help to describe the seasonality without explaining it, but economists in my experience usually regard seasonality as a nuisance, so you may not care. $\endgroup$
    – Nick Cox
    Feb 19, 2015 at 21:03
  • $\begingroup$ Be careful with seasonality. If it is there, you should model it rather than neglect it. Otherwise, funny things can happen. For example, you could generate two variables $x$ and $y$ independently of each other and add the same seasonal component $s$ to each of these. If you run a regression of $y+s$ on $x+s$, you would likely end up finding a statistically significant relationship between the two. But this is purely due to the seasonal component since you generated $x$ and $y$ independently! $\endgroup$ Feb 20, 2015 at 19:39

0

Your Answer

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