Confidence intervals r x c table?

Lorraine22 — Mon, 25 Feb 2019 15:58:26 GMT

can i get confidence intervals for each cell in a "r x c table" (based on row percent)

I have used the riskfdiff option in a 2x2 table ; but obviously this does not work on 3x2 table

data 2x2;
input Shop $ Low Total;

Outcome="2.High"; Count=Total-Low ; output;
Outcome="1.Low"; Count=Low ; output;

datalines;
Shop1 7000 15275
Shop2 2500 7500
;

proc print data=2x2;
var Shop Outcome Count;
run;

proc freq data=2x2 order=formatted;
weight Count;
table Shop * Outcome / nopercent nocol chisq riskdiff;

run;

/******************/

data 3x2;
input Shop $ Low Total;

Outcome="2.High"; Count=Total-Low ; output;
Outcome="1.Low"; Count=Low ; output;

datalines;
Shop1 7000 15275
Shop2 2500 7500

Shop3 1850 5250
;

proc print data=3x2;
var Shop Outcome Count;
run;

proc freq data=3x2order=formatted;
weight Count;
table Shop * Outcome / nopercent nocol chisq ;

run;

I've worked it out using an online calculator but can;t replicate in SAS - i would really appreciate any guidance

many thanks

Re: Confidence intervals r x c table?

PaigeMiller — Mon, 25 Feb 2019 16:08:53 GMT

As far as I know, you can only get confidence intervals of the percent in each cell using PROC FREQ if you have a one-way table.

So, you could probably turn your rxc table into a one-way table with rxc categories, and get the confidence intervals that way.

Lorraine22 — Wed, 27 Feb 2019 12:00:18 GMT

thanks for the reply Paige - I will look into your suggestion

best wishes

Lorraine