Computing confidence interval for median

[Sean_OConnor]

Folks,

I've a query which I'm hoping people may be able to lend some advice on?

At the moment I'm working on a validation study between answers an individual gives to earnings by face to face interview and what their actual earnings from administrative records. 

The information which is given by interviews is used to complie headline figures on median income of the sample and mean income.

What I would like to do is compute confidence intervals at the 95% for these headline survey figures and see do the median and mean income values taken from the administrative records fall within the confidence intervals set out from the survey. In other words are these headline figures statistically significantly different than the survey ones?

I understand how to compute upper and lower confidence intervals in SAS for means but I'm unsure how to do it for a median? From reading up it can be done, although with some slight assumptions limaitations etc.

Any input would be very welcome.

DATA Inc;
 INFILE DATALINES;
 INPUT Income_survey Income_admin;
DATALINES;
74680 11462
2076 54361
40150 76792
90681 26930
15000 15000
3600 2589
;run;

[Ksharp]

Here are two kind of CI for median,one is for normal distribution,another is free distribution.


DATA Inc;
 INFILE DATALINES;
 INPUT Income_survey Income_admin;
DATALINES;
74680 11462
2076 54361
40150 76792
90681 26930
15000 15000
3600 2589
;run;

proc univariate data=inc CIPCTLDF CIPCTLNORMA ;
var income_survey;
run;