Re: st: FW: How to make 'contract' create a row of zero frequencies

[n j cox <n.j.cox@durham.ac.uk>]

To show the issue: suppose Pauline's data looks like
this

centre   month
A          1
A          2
B          2
B          3

contract centre month, zero

will produce

centre  month _freq
A        1       1
A        2       1
A        3       0
B        1       0
B        2       1
B        3       1

Thus, -contract- gives you counts for
the Cartesian product of {A, B} and {1, 2, 3},
i.e. all possible pairs.

If you look at the code using -viewsource-
you will see that -fillin- is doing the hard work
here.

Now suppose Pauline has also centres C, D, etc.
in mind but that they are not in the dataset. How
can -contract- be forced (or, more diplomatically,
persuaded) to take those into account?

The first answwer is that -contract- has no
extrasensory ability to know what might have been
the case but isn't. Nor are there handles to specify such
details on the fly.

You need a way to work around this limitation of
-contract-. In addition to Roger's solution, another
way to do it is to add pseudo-observations

centre month
C    1
D    1

before -contract-. Then you have to make sure
that you retract your lie.

replace _freq = 0 if centre == "C" & month == 1
replace _freq = 0 if centre == "D" & month == 1

Yet another way to do it is after the -contract-.

Just add new observations like those above
and then

fillin centre month
replace _freq = 0 if _fillin
drop _fillin

In addition to the manual entry on -fillin-, see
"Filling in the gaps" in SJ 5(1) 2005.

Nick
n.j.cox@durham.ac.uk

> I have a large dataset containing information on individuals 
attending for
> screening at several centres.
>
> I am writing a program to summarise monthly attendance, to be run
> regularly by one of our administrators.  So I first 'contract' my data to
> create a new dataset of the number of attendances for each centre for
> each month.  I used the following statement:
>
> contract centre3 mrec, freq(count) zero
>
> (Centre3 = centre code, mrec = month code)
>
> I have discovered that when there are no attendances in a month at any of
> the centres, Stata completely excludes this month rather than create a
> zero count for each centre.
>
> This then completely throws out my summary statistics.  Is there any 
way I
> can force 'contract' to create a count of zero for each centre for any
> month with zero attendance at all centres?

and Roger Harbord suggested

One solution could be to use -tabcount- by Nick Cox, available from SSC.
You specify a list of values which define the categories of a variable that
are to be counted, and zero frequencies are then included. The -replace-
option makes it function much like -contract-, except zero frequencies are
retained for the specified values.

*
*   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/