Imputation of categorical missing data in SPSS

[Blain Waan]

Hello all,

I want to know a very basic thing about adjustment of missing values for categorical variables in SPSS. I have a data set containing some categorical variables. Some variables have only two categories 0 and 1. Some variables have 5 categories 0 to 4. Some variables have even 11 categories 0-10. Almost all the variables have missing values. I have 3 questions:

1) How do I impute those missing values in SPSS? Please write the procedure for me that I can understand what steps I need to follow to impute those categorical missing values.

2) I have read about hot-deck imputation while I was searching over net. Is it good for my case (note that I have some continuous variables too in my data set. I read that the background variables need to be categorical.) Again is hot-deck imputation plausible for imputing the missing values of a variable having only two categories (0 and 1)? If hot-deck in applicable is my case, how do I do this in SPSS?

3) How do I impute missing values for the continuous variables? Again if you kindly mention the steps, it will be a great help.

[Blain Waan]

I have found hot-deck useful. There are many macros that allows hot-deck in SPSS. A google search will bring many options to download these macros. It is very important to select the background variables properly. Hot-deck finds match for the missing in accordance with the background variables.

For continuous variables EM algorithm is a good option and we can even use hot-deck if we think appropriate.

[Berley]

I recommend getting the book "Advanced and Multivariate Statistical Methods" by Mertler & Vannatta. The procedures you need (as well as the underlying explanation of what to do and why to do it) are there.