09-10-2012 12:09 PM
I need your help. Is it correct to replace missing values with zero values in a data set? Might it change the results of the entire analysis?
Please answer asap
09-10-2012 12:34 PM
In general, it is not a good idea to replace missings with zeros, specifically because it can change the results of the analysis.
In some situations, a missing might be reasonably interpreted as a zero. For instance, if you have a check-box for an attribute that you might code as a 1 if present. Then coding the non-presence as a 0 would be appropriate. As a best practice, that should be a design decision rather an an analysis one.
09-10-2012 01:03 PM
thanks for replying. I have numerical values and I cannot discriminate them with presence, non-presence (1, 0). My problem is that not having values in some of my columns JMP cannot analyse a blocking as request because the count of values is unequal.
I would like to understand how to deal with these missing values considering a block as random effect.