date imputation (proc mi or data step)

[swar]

Hi. This is the structure of my data

id dt_admission dt_discharge

1 3/21/2013 3/24/2013

2 3/23/2014   3/25/2014

3 4/22/2014  4/25/2014

4                    4/24/2015

5                    5/1/2015

For the last two dates, I have to impute dt_admission based on dt_discharge. 

I have two thoughts...

imp_dt _admission = dt_discharge -(specific  number of days)  or 

imp_dt_admission = dt_discharge - (avg(dt_discharge - dt admission)) for entire data

Can anyone provide me proc mi procedure or data procedure for imputation of dates?

[thomp7050]

Out of curiosity, why use proc mi?  Why not write data steps based on the logic that you have provided?

[swar]

Yes.. That's right. But just a thought on proc mi.

[ballardw]

One might ask what the resulting set will be used for.  That may have some impact on building rules.

I can see that further information might be useful in considering building rules such as treatment information. Simple broken bones may have a notably shorter period than an open-hear surgery for instance.

And considering we had another question today about indentifying admissions (possibly multiple) of deceased persons then why admission date is missing becomes kind of interesting.