stefferrs
11-07-2007, 12:37 PM
Hi Everyone,
I'm working with a longitudinal data set that is currently split up by year (1997-2005). I need to impute values for a numeric variable (years of experience) that has quite a few missing values/invalid values. So basically, if a person has valid values in another year, I'd like to replace the invalid value with the valid value.
My question is what would be the best approach? Should I start at 2005 (the last year for which I have data) and work backward to 1997? Once I get to 1997, should I then look forward to 1998 to find valid values? If the person has multiple valid values, how should I treat that - take the average, or perhaps take the most recent value? If in the end, the person has no valid values in any of the years, should I just code it as missing?
Thanks for any advice/suggestions. Please let me know if additional info is required.
I'm working with a longitudinal data set that is currently split up by year (1997-2005). I need to impute values for a numeric variable (years of experience) that has quite a few missing values/invalid values. So basically, if a person has valid values in another year, I'd like to replace the invalid value with the valid value.
My question is what would be the best approach? Should I start at 2005 (the last year for which I have data) and work backward to 1997? Once I get to 1997, should I then look forward to 1998 to find valid values? If the person has multiple valid values, how should I treat that - take the average, or perhaps take the most recent value? If in the end, the person has no valid values in any of the years, should I just code it as missing?
Thanks for any advice/suggestions. Please let me know if additional info is required.