missing values in panel data

[sultan al shekaili]

hello

i am having missing values in my panel data. the missing values are on the dependent variables. how to best deal with this issue?

thanks
sultan

[sultan al shekaili]

in the paper frank and goyal (2009) the authors applied a technique called multiple imputation on cross-data to deal with the missing values. in the foot notes they mentioned " we use the program ICE in stat and generate five imputed data sets" page 23 .

i am trying to replicate their procedure. can you please guide me on how to do so. i have tried to for long times but couldn't .

[Carlo Lazzaro]

Sultan:
contact the corresponding author and ask if the (Stata?) codes they used are available.
Otherwise, see -mi-.

[Marcos Almeida]

I believe Carlo gave the best approach to tackle missing values in Stata, which is the multiple imputation machinery. That said, you didn't say much about your data, apart from the fact that there is missing data. For example, it is missing - sorry for the pun - the percentage, the pattern (MAR?), etc. Depending on such information, even multiple imputation will fail to produce a reliable estimation.

[sultan al shekaili]

thanks Carlo

unfortunately contact details are not available in the paper

due to the limited time i have, i proceeded and removed observations with missing values and which have less than 3 consecutive years of data.

[sultan al shekaili]

hello marco

my data is historic financial statements of sample of companies. the data missing most of the time is due not updating them by the database i use to extract these data.

[Marcos Almeida]

The "behaviour" in this case relates to eventual bias concerning the updating of Yvar. If this is surely random, this is one thing. If the firms without regular updates "behave" differently, this is quite a different thing. A panel data structure with missing values won't forbid Stata to do its work. Plainly removing observations with missing vaues may potentially bias the results, let alone the issue of the decrease in power.