Re: Data Analysis Question -- WHAT??

Posted by David Moore on March 30, 2000 at 11:56:03:

In Reply to: Re: Data Analysis Question -- WHAT?? posted by WHAT?? on March 14, 2000 at 09:47:20:

Let's see, this is the year 2000 isn't it? Doesn't anyone read the statistical literature on the subject anymore? To put it simply, if there's enough missing data for any approach to make a difference, then deletion of missing data (i.e., leaving blanks blank) will always produce biased results. Depending on why the values are missing in the first place, however, you may be able to improve your results substantially by imputing values. There are very sophisticated techniques available that may do a good job. To claim that "ANY ATTEMPT to recode the BLANK data is, as Jerry mentioned, falsifying data... " is misleading. Does this include probes by interviewers to get hesitant respondents to reveal their true attitudes? Does this include techniques that obtain the value the respondent would have provided were they inclined to do so? It would be useful to look at this from a technical, what's the least error prone approach, rather than resort to the polemical absolutes we may have received years ago as naive students.

Follow Ups:

Subject: