Dear Allstatters,
when we give a frequency table for a variable we include "missings" as a group on which we have had no information.
The missing cases are not the same for every variable in the (in this case) questionnaire. Case 1,15 and 230 are "missing cases"
for their answer on "age" but it is other respondents that did not give us information on their "education".
The respondents as a group are supposed te be representative for the population we are investigating.
So, in presenting the results we say that our group of study has this age and that education, and we add each time that
there are x(number) missing cases.
My question:
what must we do when we compute a mean for a variable? must we include the missing cases or not? the result of the mean
will differ, as the respondents that did not give us information for that variable will lower the mean if we include them.
If we omit the missing cases, are we not making a bias in the measures we are computing?
In the computing of one variable, we omit case A and B, but for another measure, we omit other cases. What is the representativity
of the group of study? Is it not better to compute every measure on the basis of the total group, including the missing cases?
I would be very grateful for every reaction.
Thanks in advance.
Erwin
Erwin Legiest - Geertrui Vandelanotte
Sint Franciesdijk 25
B-9185 Wachtebeke
Belgium
tel +32 (0) 9 345 81 70
e-mail: [log in to unmask]
|