Example 2: Split-Half
Reliability
Continuing from Example
1, suppose you want to use the questionnaire to assess the success
of a new advertisement campaign. In that case you may want to divide the
8-item scale into two halves that can be used before and after respondents
have been exposed to the new commercials. You can split the scale "in
the middle," that is, make up one sum scale of items 1 through 4,
and the other scale of items 7 through 10.
Specifying the analysis.
In the Reliability
and Item Analysis Startup Panel, click the
Split-half reliability button to display the standard variable
selection dialog. Select variables Item1-Item4 for the First
half and variables Item7-Item10 for the
Second half and then click the
OK button. Also on the Advanced tab select NO
in the Correlation matrix box.

Click the OK button on the
Startup Panel to display the Reliability Results dialog.
Reviewing the results.
Now, the Split-half reliability
button is available on the Reliability Results - Advanced tab.

Click the Split-half reliability
button to display the Split-Half
Reliability results dialog and then click on the Advanced tab.

As you can see in the summary box, the estimate of reliability based
on the Split half reliability
formula in the Introductory
Overview is practically identical to the internal consistency estimate
(Cronbach's alpha - see Example
1). The Guttman split-half reliability
reported on this dialog can be interpreted as the Cronbach's
alpha that would result if you treat the two halves of the scales
as two single items of a two-item overall scale. Again, this estimate
is very similar to the other reliability coefficients.
Click the First half means and std.
dev's button to display the Means
and Standard Deviations spreadsheet for the first half of the scale.

Finally, click the Second half means
and std. dev's button to display the Means
and Standard Deviations spreadsheet for the second half of the
scale.

It appears from the Means and Standard
Deviations spreadsheets that the means
and standard
deviations for the two halves of the scale are also comparable. Therefore,
you may use these two halves as equivalent scales to measure prejudice
against foreign-made cars.