How to forecast the ratio of y1 and y2?

[Jade_SAS]

Hi All,

    I have a data set with hierarchy City----Store, and several independent variable (var1-var3), time is daily, there is one ratio variable I'm interested to forecast (Ratio=Total #  y1/ Total # y2 at that level), I have those two variables (y1, y2) in the data. I wonder whether it's possible to forecast the ratio? Thank you!

   The data sample is as below:

Time	City	Store	Var1	Var2	Var3	Total # y1	Total # y2	Ratio(y1/y2)
	A	1
	A	2
	A	3
	B	1
	B	2
	B	3
	B	4
	C	1
	C	2
	C	5

In my understanding, we have the data of ratio (y1 and y2) at the store level and at daily frequency, but we could not use this ratio to get the weekly and city level data (which means we could not use the ratio directly for aggregation and Accumulation), we need to get the total # of y1 and total # of y2 at city level and weekly frequency, then get the ratio= total# y1/ Total# y2. Thus I don't think we can forecast the ratio at this data structure.

Please share your insight on this problem.  Thank you!

Jade

[ptimusk · Accepted solution]

This seems like an aggregate v record level ratio.

I might solve this by grouping in a summary statistic in a PROC SQL using a group by statement

like so

PROC SQL;

create table want as

select *,

sum(y1) as totalY1,

sum(y2) as totalY2,

sum(y1)/sum(y2) as RatioY1_Y2

from havedatasample

group by city, store;

quit; run;

[Jade_SAS]

Thank you!