Archive for the ‘Statistical Packages’ tag
Missing Data Analysis
When taking THE ‘Market Intelligence’ course here at Goizueta, we came across a data set that contained lots of missing values. Yes, I know that most of the data sets you see out there have some missing data or the other. But when I started looking for material on how to deal with these data points (other than simply discarding them) I was frustrated not to find anything for a manager-type audience. Most of the discussion was for statisticians. So, based on what I heard in class, here is a flow diagram that should help. Note that there are some acronyms used, but you can look them up and find out what they mean. They are not difficult.
- Various Statistical packages have automated functionality to achieve this.
- Various options available to impute data. Use business judgement.
Optional
- Run analysis with imputed data and dropped cases
- Check for any significant difference in the models
- Provide necessary details of imputation when presenting analysis
