Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: st: Carrying forward last value conditioned on time since last non-missing value- panel data
From
Nick Cox <[email protected]>
To
"[email protected]" <[email protected]>
Subject
Re: st: Carrying forward last value conditioned on time since last non-missing value- panel data
Date
Wed, 28 Aug 2013 08:51:53 +0100
I assume -testdate- is a daily date variable formatted %td or %d.
What was the previous date measured?
clonevar prevdate = testdate if !missing(glucose)
bysort id (testdate) : replace prevdate = prevdate[_n-1] if missing(glucose)
What was the previous value measured?
clonevar prevmeas = glucose if !missing(glucose)
by id: replace prevmeas = glucose[_n-1] if missing(glucose)
This is an FAQ! See
http://www.stata.com/support/faqs/data-management/replacing-missing-values/
but note the warning there on interpolation.
Replace missings if within one year
clonevar glucose2 = glucose
replace glucose2 = prevmeas if missing(glucose)
& inrange(testdate, prevdate, prevdate + 365)
The last two lines are one command.
But all that said, this seems an arbitrary procedure. Suppose a year
and a day elapses between measurements. Do you still want to use the
previous value a year ago, not the next?
For interpolation solutions, see
-ipolate- (official)
-cipolate- (SSC)
-csipolate- (SSC)
-pchipolate- (SSC)
-nnipolate- (SSC)
Nick
[email protected]
On 28 August 2013 07:03, Amit Achhra <[email protected]> wrote:
> I was wondering if anyone could help me in this. I have panel data of the form:
>
> id testdate glucose Desired output for glucose
> 1 23-Apr-09 5.9 5.9
> 1 24-Jun-09 . 5.9
> 1 23-Jul-09 6.7 6.7
> 1 24-Aug-09 . 6.7
> 1 6-Nov-09 . 6.7
> 1 22-Jan-10 . 6.7
> 1 22-Apr-10 5.8 5.8.
> 1 24-May-11 . .
> 1 3-Aug-11 6.4 6.4
> 1 28-Dec-11 . 6.4
> 1 12-Sep-12 . .
>
>
> To fill in missing values of the variable glucose, I would like to carry last non-missing value forward but only for one year ahead (i.e. only for 1 year from the last non-missing value) (see 'desired output').
> This is because I'm assuming that glucose will not have an impact on my outcome if it was measured more than a year ago.
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/faqs/resources/statalist-faq/
* http://www.ats.ucla.edu/stat/stata/