Missing value

[PhanS]

I am handling a cross-sectional data set.

I identified that a number of my variable of interests were missing values, including continued, bi-variate and ordinal data. 

I am wondering if I can replace the missing values by its mean or median.     

And I want to know if there is a cap for allowing the replacing its value, e.g., missing values must less than 10% or so.  

Anyone can help much appreciated.

Phan S.

[stat_sas]

Hi,

Missing values can be replaced with various statistics using proc stdize. Below is an example replacing missing values with median.

Defining a cap would be based on your analysis. You can flag variables containing a certain percentage of missing values for imputation.

proc stdize data=have reponly method=median out=imputed;
var a b c; /* Assuming a, b and c are 3 numeric variables */
run;

[PhanS]

Hello,

I am sorry, I mean I give the solution (credit) to you, bu accidentally check to Reeza. 

Thank you for you code.

Phan S. 

[Reeza]

@PhanS You can change that to @stat_sas, just select theirs instead.

[stat_sas]

Hi,

I am glad you have the solution. I am also  learning from @Reeza's posts 

[Reeza · Accepted solution]

Here's an inefficient macro that will cap the outliers. 

https://gist.github.com/statgeek/31316a678433a1db8136

PROC STDIZE is a better option for replacing with median/mean. You can also look into PROC MI, multiple imputation to impute missing data.