Hi 1-4-the-road
Larry the husband here. Thanks a lot for the explanation. It makes sense and backs up what drunkie said. i switched to 7 day trailing averages and the trend looks more realistic. I'm going to play around with it some more and see what kind of projections I come up with.

Best wishes.

Hi Larry

I'd (mildly) caution against fitting an exponential through a trailing average - as you are in essence double smoothing. As long as you do it on the daily averages it probably won't make much difference I suppose, but averaging over a longer period may give you unrealistic trendlines.

Good to know, I WAS going to tool around with 2 and 3 day averages.

Yes I was suggesting trailing average as a trendline, not source for another trendline. On the other hand weekly totals is good source for trendline. The reason daily is not such a good source is day to day variations and especially AF days which are very good thing but not for trendlines.

To be honest, exponential trendline in these series does not seem to fit, no matter what you do.

I'd spend my energy on other things like median number of drinks or total number AF days, etc, if number of drinks does not appear to go where you want it to.

You could also consider outlier detection. I.e. it was vacation or new year or someone's birthday and whereas normally you'd get 5 there is no natural stop and you get 10. Then cut everything at 5 before averaging.

When they do experiments on rats they don't have to give them same rights as we humans have. Rats can't go to store if pedal stopped working and they don't have family coming to town with bags of liquor.

Good ideas, I'll flag the outliers on trigger days. thanks again.