Overview (RELIABILITY command)
RELIABILITY
estimates reliability statistics for the components of multiple-item
additive scales. It uses any one of five models for reliability analysis
and offers a variety of statistical displays. RELIABILITY
can also be used to perform a repeated measures
analysis of variance, a two-way factorial analysis of variance with
one observation per cell, Tukey’s test for additivity, Hotelling’s T-square test for equality of means in repeated
measures designs, and Friedman’s two-way analysis of variance
on ranks. For more complex repeated measures designs, use the GLM
procedure (available in the Advanced
Models option).
Options
Model Type. You can specify any one of
five models on the MODEL
subcommand.
Statistical Display. Statistics available on the STATISTICS
subcommand include descriptive statistics, correlation and covariance
matrices, a repeated measures analysis of variance table, Hotelling’s T-square, Tukey’s test for additivity,
Friedman’s chi-square for the analysis of ranked data, and
Cochran’s Q.
Computational Method. You can force RELIABILITY
to
use the covariance method, even when you are not requesting any output
that requires it, by using the METHOD
subcommand.
Matrix Input and Output. You can read data in the form
of correlation matrices and you can write correlation-type matrix
materials to a data file using the MATRIX
subcommand.
Basic Specification
The
basic specification is VARIABLES
and a variable list. By default, RELIABILITY
displays the number of cases, number of items, and Cronbach’s
alpha. Whenever possible, it uses an algorithm that does not require
the calculation of the covariance matrix.
Subcommand Order
-
VARIABLES
must be specified first. - The remaining subcommands can be named in any order.
Operations
-
STATISTICS
andSUMMARY
are cumulative. If you enter them more than once, all requested statistics are produced for each scale. - If you request output that is not available for your
model or for your data,
RELIABILITY
ignores the request. -
RELIABILITY
uses an economical algorithm whenever possible but calculates a covariance matrix when necessary (see METHOD Subcommand (RELIABILITY command)).
Limitations
- Maximum
1
VARIABLES
subcommand. - Maximum 1
SCALE
subcommand. - Maximum 10,000 variables on the
VARIABLES
subcommand. - Maximum 10,000 variables on the
SCALE
subcommand.