Tuesday, May 22, 2012

Missing values and missing values and missing values

It used to be common practice to code missing values with a numeric substitute that you would know would be a missing value. I'm not sure what the origins of the practice are, but I'm fairly certain it had to do with limitations in how the computer stored data. Fortunately, our systems have evolved to a point that we don't have to use this practice any more. Instead, we can place a blank in the data and simply tell the computer to interpret that as missing. Unless you're using SPSS. In which case, you still need to define a number to be the "missing value." In the data I'm currently working on, the options for missing data are 888, 999, 8888, 9999, 88888, and "n/a". Is it any wonder that no one with any credibility uses SPSS anymore?

No comments:

Post a Comment