|
|
Dear SPSS experts,
We hava a dataset with 95 items, formated in SPSS as 95 cases.
For each of these cases we have results at 5 points in time for a repeated frequency-based
measurement (represented by 5 variables in SPSS). Now we want to integrate these 5 (correlating)
results (for each item/case sparately) into one value of a new variable (i.e., a general frequency
estimate).
Q1: Instead of just using the arithmetic mean of the five values we want to use m-estimators, since
these five values sometimes differ obviously (e.g., 456, 789, 1490, 5234, and 200 for one item). As
SPSS offers 4 different m-estimators we are now faced with the question: Which one suits best for
our purpose?
Q2: Another question: Would be the median an adequate estimate in this case?
Q3: Finally: Do you know a scientific reference adressing this question and supporting your
recommendation?
Thanks in advance and happy new year
Joerg-Henrik
|
Locked
