Creating a Last Observation variable

[ejf ejf]

I have a large dataset with 112 variables with each giving each person's weight at attendance1, attendance2 etc. If the individual did not attend that week then the value for that week is missing as they were not weighed.

I want to create a Last Observation Carried Forward where at 112 weeks I have a value for each person that is equal to the last time they were weighed.

Does anyone know how I go about this? I can't figure it out.

[Carlo Lazzaro]

Ejf (is it your real name?):
LOCF is, statistically speaking, one of the several forms of quantitative evil.
Besides, if your study follows patients who should lose weight, those missed at follow-up may well be those less interested or compliant: hence, your missingness is informative.
In addition to the wide literature on this topic, I would also point you to the following online resource: http://www.missingdata.org.uk/. That website is maintained by Jonathan Bartlett, whose posts appeared from time to time on this forum until some months ago.

[Nick Cox]

While agreeing with all of Carlo's comments -- including the detail on your name (ejf ejf is implausible as given and family name -- please respect forum practice on this point) -- I'd point out that there are many interpolation methods likely to be better than this one.

https://www.statalist.org/forums/for...-interpolation points to the most versatile interpolation command I know of for Stata.

I'm not sold on multiple imputation as a way of filling gaps in time series unless the methods used respect their dependence structure. Equally any interpolation method carries an obligation to check how much difference it makes. I'd compare original data and the results of at least two different interpolation methods

Just treating gaps as part of the data is not outrageous either.