Should I detrend data when doing a hypothesis test? It's daily sales data. One group with days of rain, one with days without. Taken from a group of about 100 days from the same time period ie Feb 2-may 5. I'm very new to this. Any help is greatly appreciated.
$\begingroup$
$\endgroup$
0
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
1
I would try both ways because you don't really know how successful your "detrending" will be. If you get the same answer both ways, that is great news because it means you likely have a really strong difference. If not, you need to examine the data more closely to see which has fewer issues. It would also be helpful if you had data from the same time period but a year before (assuming all other things stayed constant).
You didn't say how you want to do this but there are easy ways to decompose time-series data and end up with something like:
-
$\begingroup$ Thanks! I took the samples from the same quarter of sales. I wasn't the manager last year, so I stayed away from that. I figure, if detrended then being all from the same pool I can assume equal variances and do a simple t test. $\endgroup$– BrandonCommented May 30, 2017 at 22:15