Calculate sensitivity and specificity without generating 2 by 2 table

[DeepakSwain]

Hi there,

I am trying to calculate sensitivity and specificity of a test (e.g. test2) with reference to a standard test (e.g. test1). I am interested to know a way where I can calculate sensitivity and specificity without multiple proc freq steps.  A sample of my data is gievn below: 

data sensitivity_study;
input test1 test2;
datalines;
1 1 
1 0 
0 1 
0 0
1 1 
1 0 
0 0 
0 0
0 0 
0 0 
;
run;

Thank you in advance for your kind reply. 

With Regards,

Deepak

[Reeza]

This is the same as this question, essentially.

https://communities.sas.com/t5/SAS-Procedures/Getting-PROC-FREQ-Report-Output-With-Another-Format-an...

[Ksharp · Accepted solution]

This could give you a start .



data sensitivity_study;
input test1 test2;
datalines;
1 1 
1 0 
0 1 
0 0
1 1 
1 0 
0 0 
0 0
0 0 
0 0 
;
run;
proc sql;
 select test1,sum(test1=test2)/count(*) as sensitivity,
   case when test1=0 then 'specificity' else 'sensitivity'  end as name
  from sensitivity_study
   group by test1;
quit;

[DeepakSwain]

Hi Ksharp,

Great solution as I have been looking for without generating 2 by 2 table. 

Thanks.

With Kind regards,

Deepak

[DeepakSwain]

Hi Ksharp,

Is it possible to extend this code to calculate PPV and NPV. 

Thank you in advance for your kind reply. 

Regards,

Deepak

[Ksharp]

Technically , it could be . Give us formula about them .

Or you could change 

sum(test1=test2)/count(*)

into

sum(test1 ne test2)/count(*)

 and see whether it is what you want.

[DeepakSwain]

Hi Ksharp,

I tried the following sas code but could not generate the desired result.

data sensitivity_study;
input test1 test2;
datalines;
1 1 
1 0 
0 1 
0 0
1 1 
1 0 
0 0 
0 0
0 0 
0 0 
;
run;

proc sql;
 select test1,sum(test1 ne test2)/count(*) as ppv,
   case when test1=0 then 'NPV' else 'PPV'  end as name
  from sensitivity_study
   group by test1;
quit;

Can you kindly help me further. 

Positive Predictive Value =True Positive/(True Positive + False Positive)

Negative Predictive Value=True Negative/(True Negative + False Negative) Thank you in advance for your kind reply. Regards, Deepak

[Reeza]

The logic is to write a sum statement with the conditions inside the sum statement that meet your requirements, ie true positive is test1=1 and test2=1.

The first link I posted demonstrates this a bit. You should be able to extend the logic already provided to your new scenario. Please try and post what you've tried and isn't working if you need further help. 

[Ksharp]

OK. Just change GROUP variable into TEST2.



data sensitivity_study;
input test1 test2;
datalines;
1 1 
1 0 
0 1 
0 0
1 1 
1 0 
0 0 
0 0
0 0 
0 0 
;
run;

proc sql;
 select test2,sum(test1=test2)/count(*) as ppv,
   case when test2=0 then 'NPV' else 'PPV'  end as name
  from sensitivity_study
   group by test2;
quit;

[DeepakSwain]

Hi Ksharp,

Kindly accept my apology for the delayed reply. Your tips has worked nicely and has provided me a very simple way to calulate Sensitivity, Specificity, PPV, NPV without creating 2 by 2 table. 

Thanks a lot for your honest effort to help me. 

Appreciated.

Regards,

Deepak