observations with N/A reducing the sample

[rob86]

Hello (accidentally posted this in the econometrics discussions forum so sorry about repost)

I have a sample of 500 firms with around 50 explanatory variables. Out of the around 30000 observations there are 2000 N/A's. When I run the regression the sample size is reduced by any firm with has an N/A in any of the explanatory variables which leaves around 200 of the 500 firms. Is there any way to keep the whole sample of 500 firms when some of the observations are N/A's? or will an N/A in any of the explanatory variables always remove the firms from the sample?

Thanks

[startz]

Hello (accidentally posted this in the econometrics discussions forum so sorry about repost)

I have a sample of 500 firms with around 50 explanatory variables. Out of the around 30000 observations there are 2000 N/A's. When I run the regression the sample size is reduced by any firm with has an N/A in any of the explanatory variables which leaves around 200 of the 500 firms. Is there any way to keep the whole sample of 500 firms when some of the observations are N/A's? or will an N/A in any of the explanatory variables always remove the firms from the sample?

Thanks

The latter. It's called listwise deletion.

[rob86]

thanks for second answer in two days startz, anyone have an opinion on using Hot deck imputation when <4% of observations are missing? would this have any major negative impacts?