topic Re: imputing categorical variables in SAS Programming

imputing categorical variables

Solly7 — Mon, 14 Jun 2021 09:53:55 GMT

Hi, i need help in imputing categoriacla variables.. i tried below code but im not sure which value to use on the 'method' option...i have a variable called 'marital status' consistin of values 'single, married,widowed'...i need to do something like mode imputation

proc stdize data=data out=Imputed
oprefix=Orig_ /* prefix for original variables */
reponly /* only replace; do not standardize */
method=; /*ABW, AGK, AHUBER, AWAVE, EUCLEN, IN,IQR, LEAST, MAD, MAXABS, MEAN, MEDIAN, MIDRANGE, RANGE, SPACING, STD, SUM, UST*/
var Marital_status; 
run;

Re: imputing categorical variables

PaigeMiller — Mon, 14 Jun 2021 11:09:44 GMT

From the PROC STDIZE documentation:

VAR Statement

VAR variable;

The VAR statement lists numeric variables to be standardized. If you omit the VAR statement, all numeric variables not listed in the BY, FREQ, and WEIGHT statements are used.

So PROC STDIZE cannot be used on character variables.

To determine the mode of a class variable, you can do this in PROC FREQ and then once you have found the mode, replace missings in your data set with the calculated mode.

Re: imputing categorical variables

Ksharp — Mon, 14 Jun 2021 11:18:44 GMT

If I was right PROC STDIZE can't handle category variable.
or maybe @Rick_SAS know more .

Re: imputing categorical variables

Solly7 — Mon, 14 Jun 2021 11:58:07 GMT

I fully agree, i tried below code...i initially had 24000 rows before applying below code..but thereafter executing the below code my data multiplied to 505000 rows..kindly assist @Rick_SAS

proc mi data=final_data seed=1305417 out=outex7;
   class Existing_cover Marital_status Lead_type;
   fcs discrim(Existing_cover Marital_status Lead_type/details) reg(height_new/details);
   var Existing_cover Marital_status Lead_type height_new;
run;

Re: imputing categorical variables

Rick_SAS — Mon, 14 Jun 2021 12:59:26 GMT

You need to decide what kind of imputation you want. Originally, you asked about PROC STDIZE, which performs single imputation: missing values of numerical variables are replaced by a value or by a statistic such as the mean or median. But now you are talking about PROC MI, which performs multip[le imputation.

A paper by Paul Allison discusses the various imputation methods for various assumptions of missingness.

Re: imputing categorical variables

Sajid01 — Mon, 14 Jun 2021 13:01:56 GMT

Hello
Imputing character variables has been discussed in this forum topic. Suggest have a look there
https://communities.sas.com/t5/SAS-Programming/Imputing-missing-values-of-character-variables-in-a-large-data/td-p/465481