I have a simple ols equation where the independent variable is total production in a month. I seasonally adjust the data with an x12 and run the ols and get my estimates. The strange thing is, if I change the independent variable to average production per day by dividing by the number of days in the month, and then seasonally adjusting the series my R-squared goes up by almost 0.1. Does anyone have any idea why that might be?
6
-
1$\begingroup$ If production depends on the amount of time invested in it (which is a reasonable assumption), then the result is as expected. You are reflecting a relevant feature in the second model, but you are not reflecting it in the first one; hence, no wonder that the second model has a higher $R^2$. That should happen regardless of seasonality. However, seasonality may mess things up. $\endgroup$– Richard HardyCommented Jun 16, 2015 at 19:37
-
$\begingroup$ I don't follow you. The same information is in both series just presented a difference way. $\endgroup$– rinnovoCommented Jun 16, 2015 at 19:39
-
$\begingroup$ Well, one of us is too quick: either I did not quite get the setting right, or you did not quite get the idea behind my argument. Some more time spent reflecting should help :) $\endgroup$– Richard HardyCommented Jun 16, 2015 at 19:41
-
$\begingroup$ Why do you say the first one doesn't reflect time invested? Shouldnt the seasonal adjustment correct for the number of days in the month? $\endgroup$– rinnovoCommented Jun 16, 2015 at 19:52
-
$\begingroup$ If it does, then my comment is irrelevant. I do not know exactly what x12 does. $\endgroup$– Richard HardyCommented Jun 16, 2015 at 20:03
|
Show 1 more comment