Number of observations for each category when regressing

[Nicole Wardrop]

Hi,

I've got a very long code that runs a bunch of regressions and writes the output into CSV files. I want to include the number of observations for each category when I run a regression for example:

Model 1
Factor		n	Beta coefficient
Education level		n = 649	test of trend: p = 0.10
	Less than Year 10	28	Reference
	Year 10 or Year 11	75	1.8
	Year 12 or equivalent	177	2.5
	Trade/ Certificate	389	2.8
	Bachelor degree	247	3.3
	Postgraduate	106	3.8

I've used e(N) to get the overall number of observations for the regression (n = 649):

regress age i.education
local n = e(N)

However, the number of observations for each category currently come from just tabulating education which does not take into account the records eliminated due to missing data when running the regression. Hence the individual categories do not add up to 649.
I know I could do the following to get a table with the right numbers for the example above:

tab education if age != ., matcell(tabx)

But in reality I have a long list of variables stored in local macros that are included in the regression and I don't want to have to unpack it and manually write an if statement for each one. Is there a way to do this? Please let me know if anything above isn't clear.

Thanks,
Nicole

[Nicole Wardrop]

I've solved my problem but I'll leave it up in case anyone else wants an answer.

local outcome "age" //dependent variable
local factor "education" //independent variable
local adj "sex weight height income" //extra variables included in regression
local adjc : subinstr local adj " " ",", all //adding commas to the variable list to fit in with missing() syntax
tab `factor' if !missing(`outcome', `adjc'), matcell(tabx)

Which outputs tabx which is a matrix with the number of observations for each category of the independent variable included in the regression.

[Richard Williams]

After each regression, you could do something like

tab education if e(sample)

That will limit the analysis to the cases that were used in the regression.

Another nice command for some purposes is estat sum, e.g

reg y x1 x2 x3
estat sum

"estat summarize summarizes the variables used by the command and automatically restricts the sample to the estimation sample; it also summarizes the weight variable and cluster structure, if specified."

For more info, type

help estat summarize

[Nicole Wardrop]

Great! Thanks Richard. e(sample) is exactly what I was looking for.