Quartz | Level 8

## Sensitivity and Specificity calculations using Bootstrapped dataset

Hi SAS experts,

I have a data set with  N= 2196 observations, variables PIPE1, PIPE2, PIPE3, PIPE4, PIPE5, PIPE6, PIPE_Score,  PIIE_Triage and an outcome variable (Outcome, values 0 for No and 1 for Yes). I need help with calculating the sensitivity and specificity for predicting the outcome based on the PIPE1 using a bootstrapping data.  I calculated the sensitivity and specificity for the original data set but I need help with calculating the  sensitivity and specificity for the bootstrapped dataset. How can I achieve this? Any help with this is much appreciated.

Thank you so much in advance!

SM

PROC IMPORT OUT= WORK.Flu
DATAFILE= "\\Desktop\Flu.xlsx"
DBMS=EXCEL REPLACE;
RANGE="Sheet1\$";
GETNAMES=YES;
MIXED=NO;
SCANTEXT=YES;
USEDATE=YES;
SCANTIME=YES;
RUN;

proc freq data=Flu order=data;
tables Pipe1*Outcome / senspec;
run;
*Bootstrapping;
proc surveyselect data=flu NOPRINT seed=123456
out=fluBootout
method=urs
samprate=1
reps=1000;
run;
1 ACCEPTED SOLUTION

Accepted Solutions
SAS Super FREQ

## Re: Sensitivity and Specificity calculations using Bootstrapped dataset

On top of the below program, please also take a look at :
https://blogs.sas.com/content/tag/bootstrap-and-resampling/

Here's the program (I have made my own dataset "Flu" in the first line of code) :
And maybe you want to reconsider your choice of "1" for samprate.

The procedure treats the value 1 as 100% instead of 1%.
If you want 1%, use
SAMPRATE=0.01 .

data Flu; set sashelp.class; Pipe1=Sex; Outcome=Sex; run;

proc freq data=Flu order=data;
tables Pipe1*Outcome / senspec;
run;

*Bootstrapping;
proc surveyselect data=Flu NOPRINT seed=123456
out=fluBootout
method=urs
samprate=1
reps=1000;
run;

ods trace off;
ods exclude CrossTabFreqs;
ods output SenSpec=work.SenSpec;
proc freq data=fluBootout order=data;
by Replicate;
weight NumberHits;
tables Pipe1*Outcome / senspec;
run;

title; footnote;

title 'Bootstrapped Sensitivity';
PROC MEANS data=work.SenSpec mean median stderr stddev;
where Statistic='Sensitivity';
var Estimate;
run;

title 'Bootstrapped Specificity';
PROC MEANS data=work.SenSpec mean median stderr stddev;
where Statistic='Specificity';
var Estimate;
run;

title; footnote;
/* end of program */

Koen

3 REPLIES 3
SAS Super FREQ

## Re: Sensitivity and Specificity calculations using Bootstrapped dataset

On top of the below program, please also take a look at :
https://blogs.sas.com/content/tag/bootstrap-and-resampling/

Here's the program (I have made my own dataset "Flu" in the first line of code) :
And maybe you want to reconsider your choice of "1" for samprate.

The procedure treats the value 1 as 100% instead of 1%.
If you want 1%, use
SAMPRATE=0.01 .

data Flu; set sashelp.class; Pipe1=Sex; Outcome=Sex; run;

proc freq data=Flu order=data;
tables Pipe1*Outcome / senspec;
run;

*Bootstrapping;
proc surveyselect data=Flu NOPRINT seed=123456
out=fluBootout
method=urs
samprate=1
reps=1000;
run;

ods trace off;
ods exclude CrossTabFreqs;
ods output SenSpec=work.SenSpec;
proc freq data=fluBootout order=data;
by Replicate;
weight NumberHits;
tables Pipe1*Outcome / senspec;
run;

title; footnote;

title 'Bootstrapped Sensitivity';
PROC MEANS data=work.SenSpec mean median stderr stddev;
where Statistic='Sensitivity';
var Estimate;
run;

title 'Bootstrapped Specificity';
PROC MEANS data=work.SenSpec mean median stderr stddev;
where Statistic='Specificity';
var Estimate;
run;

title; footnote;
/* end of program */

Koen

Quartz | Level 8

## Re: Sensitivity and Specificity calculations using Bootstrapped dataset

Can we get the 95% CIs for the bootstrapped sensitivity and specificity?
SAS Super FREQ

## Re: Sensitivity and Specificity calculations using Bootstrapped dataset

@sms1891 wrote:
Can we get the 95% CIs for the bootstrapped sensitivity and specificity?

Of course, you can!

Look here and you will find out how to do so ...

Compute a bootstrap confidence interval in SAS
By Rick Wicklin on The DO Loop August 10, 2016

https://blogs.sas.com/content/iml/2016/08/10/bootstrap-confidence-interval-sas.html

Koen

Discussion stats
• 3 replies
• 833 views
• 0 likes
• 2 in conversation