Adam Chapman
> > I'm trying to calculate a running cumulative total for
> observations over
> > the previous 12 months (cum12month). My data looks like:
> >
> > # Obs Month Cum12month
> > 2 1 2
> > 4 2 6
> > 3 3 9
> > 1 4 10
> > 2 5 12
> > 4 6 16
> > 2 7 18
> > 1 8 19
> > 5 9 24
> > 3 10 27
> > 2 11 29
> > 4 12 33
> > 5 13 36
> > 2 14 34
> > 3 15 34
> > 2 16 35
> > 4 17 37
> > etc
> >
Kieran McCaul
> sort month
> gen cum12month=month[1]
> replace cum12month=month[_n]+ cum12month[_n-1] if _n>1
That's not what Adam is looking for, setting aside
the detail that he is summing Obs, not Month.
If the cumulative sum were what was wanted,
we could get it directly by
gen Cum12month = sum(Obs)
However, that is also the correct solution -- for the
first twelve months -- or more precisely coorect if we are
treating months before the start of the data as if they
contained zero values, as does the calculation behind
Adam's worked example.
But we certainly need to fix it for months 13 on. The
sum over the last twelve months is the
cumulative sum so far
MINUS the
cumulative sum 12 months ago
so we
replace Cum12month = Cum12month - Cum12month[_n-12] if month > 13
Note that this avoids a loop over observations, always expensive.
Note also that you can do a lot of arithmetic within the subscript.
Nick
n.j.cox@durham.ac.uk
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/