OT: Clustering versus nesting

[Bruce Weaver]

OT, because this is not specific to SPSS.

In another thread, I noticed that "clustering" and "nesting" were being treated as synonymous--and I have seen this elsewhere many times.  I was reminded of the time I posted something to one of the sci.stat.* groups describing a situation where time points (or occasions of measurement) were clustered within subjects.  But I used the word "nested" rather than clustered.  This caused confusion because some readers interpreted "nested" in the way that it is used for describing nested ANOVA designs.  To illustrate:

A and B crossed - every combination of A and B occurs
A1: B1 B2 B3
A2: B1 B2 B3
A3: B1 B2 B3

B nested within A - each level of B appears in only one level of A
A1: B1 B2 B3
A2: B4 B5 B6
A3: B7 B8 B9

Clustering: E.g., time points (T) clustered within subjects (S)
S1: T0 T1 T2
S2: T0 T3 T4
S3: T0 T2 T5
S4: T6 T7 T8

The last of these illustrates the situation I was talking about.  Notice that some levels of T may appear in more than one level of S; but other levels of T may not appear in all levels of S.  But there is still clustering of measurement occasions within subjects.

So rather than treating these terms as synonymous, I think it is more appropriate to say that nesting is a specific type of clustering in which each level of the nested variable appears in only one level of the other (higher order) variable.  

HTH.