Output multiple hypothesis tests into nice looking table

[James Lilleker]

Dear Statalisters,

I am tearing my hair about here. This seems like a very simple issue but I just can't find anything that works. Sorry if it is a bit of a stupid question.

What I have is this:

I wish to cross-tabulate multiple variables (coded as present or absent) according to the participant being antibody positive or negative and then do a hypothesis test. I have added in the cc command as I wish to present odd's ratios as well as p values:

foreach var in VAR1 VAR2 VAR3 VAR4{
tab `var' ANTIBODY, chi2
cc `var' ANTIBODY
}

I have wasted a lot of time copying and pasting in to MS word and making errors along the way. Is there a way to ask Stata to make a nice looking table like this:

Variable ANITBODY POS ANTIBODY NEG p OR
VAR1 n (%) n (%) x.xxx x.xx-x.xx
VAR2 n (%) n (%) x.xxx x.xx-x.xx
VAR3 n (%) n (%) x.xxx x.xx-x.xx
VAR4 n (%) n (%) x.xxx x.xx-x.xx
VARX n (%) n (%) x.xxx x.xx-x.xx


Thank you in advance...

James

[Daniela van Santen]

HI James, have your tried "tabout"?

it can make 2x2 tables and export it to excel. Don't know if this helps.

[James Lilleker]

Hi,
I have tried to use tabout, but don't seem to be able to make it perform individual tests for each variable and then combine into a single table.

James

[Elena Paltseva]

Hi James,

The below is not elegant, but I think it would work. Something similar worked for me...
Please ask if anything is unclear, and obviously I apologize if the mistakes are still there...

Best,
Elena

capture erase tabstat.dta
*cleans STATA datafile to be used
tempname hdle

postfile `hdle' str32(var) Variable ANITBODY_POS AP% ANTIBODY_NEG AN% p OR using tabstat
*opens new STATA datafile for writing

foreach var of varlist VAR1 VAR2 VAR3 VAR4 {
quietly {
count (`var') if `var'!=""
scalar Total=r(N)
*calculated the number of non-missing observations for `var' and saves in into scalar called Total

count ( `var') if ANTIBODY=="positive"&`var'=="present"
scalar AP=r(N)
*calculated the number of observations for `var'="present" if ANTIBODY=="positive" and saves in into scalar called AP

count ( `var') if ANTIBODY=="negative"&`var'=="present"
scalar AN=r(N)
*same for AN

scalar shareAP=AP/Total*100
scalar shareNP=NP/Total*100
*calculates percentages of the split

tab `var' ANTIBODY, chi2
g CHI2=r(chi2)
*saves Chi2 statistics into the scalar CHI2

cc `var' ANTIBODY
g OR=r(or)
*saves or statistics into the scalar OR

post `hdle' ("`var'") (AP) (shareAP) (AN) (shareAN) (CHI2) (OR)
*posts this to a Stata datafile

drop AP shareAP AN shareAN CHI2 OR
*clears the scalars for the next round (in principle, not needed)
}
}

postclose `hdle'
*closes posting to STATA datafile

use tabstat, clear

* Saves STATA data to spreadsheet
xmlsave "Table 1", doctype(excel) replace

[James Lilleker]

Hmm... Could get that to run.

I did some debugging and have ended up with this, but I just get a syntax error:

tempname hdle
tempfile tables1

postfile `hdle' VariableName ANITBODYPOS APpercent ANTIBODYNEG ANpercent p OR using `tables1'


foreach var of VAR1 VAR2 VAR 3 {
count (`var') if `var'!=.
scalar Total=r(N)


count (`var') if antibody==1 & `var'==1
scalar AP=r(N)

count (`var') if antibody==0 & `var'==1
scalar AN=r(N)


scalar shareAP=(AP/Total)*100
scalar shareNP=(NP/Total)*100


tab `var' antibody, chi2
scalar CHI2=r(chi2)


cc `var' MUP4401
scalar OR=r(or)


post `hdle' ("`var'") (AP) (shareAP) (AN) (shareAN) (CHI2) (OR)

}


postclose `hdle'


use `tables1', clear

xmlsave "Table 1", doctype(excel) replace

[Elena Paltseva]

I think I see at least one mistake in my (and then your) code: all lines that use

count (`var') if condition

should become

count if `var'!=. condition

(as you count observations only).

Sorry, I first wrote it differently, with egen function, and then forgot to change. Does it help?

[Elena Paltseva]

Correction:

count (`var') if condition

should become

count if condition

[James Lilleker]

finally got there... had to change foreach var *of* to *in*

thanks for you help

James

[Christophe Kolodziejczyk]

Have you tried estpost tabulate from the SSC's estout package by Ben Jann? I think you can get all the wanted results from a single command and make a table with estout or esttab.