1: Back in the middle
20th century, reliability and validity were discussed as a property
of the scale (i.e., the Osborne Obsequiousness Scale is reliable and
valid). Modern APA and other guidelines recommend that we talk about
reliability and validity as the property of samples, not instruments.
However, some instruments do tend to appear more reliable across samples,
and some less so. Hence the need for replication.
[return]
2: Although neither practice seems
to have been adopted widely in the literature, we are familiar with.
[return]
3: This implies that there is a distribution
of split-half coefficients based on different splits, and that alpha
is the mean of all these splits. This is an interesting idea that
many of us miss, as we focus just on the one number we calculate.
[return]
4: Most people
would agree this statement is “self-evident”—a
nice way of saying “well, duh!” but it is surprising
that this simple “well, duh!” sentiment is so problematic
in practice.
[return]
5: We know that this sample of N=656
is relatively small to consider its results close to our “population”
estimates, but we will do so anyhow for the purpose of this example.
[return]
6: Note: The characters &ss represent
the value of the input parameter for the macro. Thus, when we are
using N=50, our output data
set will be orig_alpha50. This is true for the remaining examples.
[return]