topic Re: Calculate weighted proportion, weighted sensitivity, and weighted specificity in Statistical Procedures

Calculate weighted proportion, weighted sensitivity, and weighted specificity

Barkat — Mon, 10 Oct 2022 23:53:52 GMT

After data analysis, I have created a table like below. I was asked to calculate weighted proportion, weighted sensitivity, and weighted specificity of each type. I googled a lot but no success. Please help. If not possible in SAS, is there any way to calculate manually, at lease.

Type	Count_TotalPop	Count_SurveyPop	Sensitivity	Specificity	Proportion_TotalPop	Proportion_SurveyPop
A	4000	20	79	67	0.23	0.17
B	3500	30	100	69	0.20	0.26
C	3000	25	90	78	0.17	0.21
D	2000	20	100	100	0.11	0.17
E	5000	22	75	55	0.29	0.19

Re: Calculate weighted proportion, weighted sensitivity, and weighted specificity

Ksharp — Tue, 11 Oct 2022 11:59:09 GMT

calling @StatDave

data have;
 set sashelp.heart;
 keep status bp_status sex;
run;

proc freq data=have noprint;
table bp_status*status*sex/out=temp list;
run;

proc freq data=temp;
by bp_status;
table status*sex/senspec ;
weight COUNT;
run;

Re: Calculate weighted proportion, weighted sensitivity, and weighted specificity

SAS_Rob — Tue, 11 Oct 2022 12:24:44 GMT

One easy way to do this would be to use Proc SURVEYFREQ with the WEIGHT statement and the SENSPEC option on the TABLES statement.

Re: Calculate weighted proportion, weighted sensitivity, and weighted specificity

Barkat — Tue, 11 Oct 2022 12:56:41 GMT

I like your idea, but I was wondering, WEIGHT of which variable? Would you please give me an example?

Re: Calculate weighted proportion, weighted sensitivity, and weighted specificity

SAS_Rob — Tue, 11 Oct 2022 13:30:07 GMT

I am not sure I understand what you mean when you ask "weight of which variable?" Typically the weight is part of the sample design. I would suggest going back to the person who asked for the weighted statistics and ask them what they want you to weight the analysis by.

As far as an example you could look at the Getting Started example in the SURVEYFREQ statement. Note how the data has a variable in it called SamplingWeight which was part of the design.

SAS Help Center: Getting Started: SURVEYFREQ Procedure

In your case since you only have weights and not a complex survey design, you would remove the CLUSTER and STRATA statements.

proc surveyfreq data=SIS_Survey;

tables SchoolType * Response / senspec;

weight SamplingWeight;

run;

Re: Calculate weighted proportion, weighted sensitivity, and weighted specificity

ballardw — Tue, 11 Oct 2022 15:27:37 GMT

Weighted how? What weight is to be applied where?

@Barkat wrote:

After data analysis, I have created a table like below. I was asked to calculate weighted proportion, weighted sensitivity, and weighted specificity of each type. I googled a lot but no success. Please help. If not possible in SAS, is there any way to calculate manually, at lease.

Type Count_TotalPop Count_SurveyPop Sensitivity Specificity Proportion_TotalPop Proportion_SurveyPop

A 4000 20 79 67 0.23 0.17

B 3500 30 100 69 0.20 0.26

C 3000 25 90 78 0.17 0.21

D 2000 20 100 100 0.11 0.17

E 5000 22 75 55 0.29 0.19