Reliability and Item Analysis

Standard reliability and item analysis for scale (test) development; applies the classical testing model to the construction of sum scales. STATISTICA computes reliability estimates (Cronbach, split-half), correction of correlations for attenuation, and provides a number of other options useful in scale (test) construction.

Element Name Description
General
Detail of computed results reported Specifies the detail of computed results reported. If Minimal results is requested, only the summary reliability statistics will be displayed. If Comprehensive results is selected, STATISTICA will report detailed item-level statistics. If All results is requested, various descriptive statistics and graphs are reported as well.
Creates correlation matrix Select whether to compute a correlation matrix for all items in the scale; the Tetrachoric correlations options can be used for dichotomized (i.e., 0/1, Yes/No) items.

 To compute Tetrachoric correlations, all values in the selected variables must be 0 (zero) or 1 (one). However, because Tetrachoric correlations may overestimate the shared variance of the items in the scale, they are not recommended for most standard analyses (consult the Electronic Manual for additional details).
Creates multiple R Select this option to calculate the multiple correlation (between each item and the sum of all others in the scale). This calculation may increase processing time if the scale contains many items (i.e., if many items were selected for the analysis). This option is ignored if no correlation matrix is computed.
MD Deletion Missing data can be deleted Casewise or missing data can be substituted by the means for the respective variables.
Number of Items, Attenuation
Creates required k items Creates the required number of items to add to the scale to achieve a desired reliability (see parameter Desired reliability).

 The computations for this option can only be evaluated if the reliability of the current scale is better than 0 (zero).
Desired reliability Specifies the desired reliability to compute the required number of items (of the same quality) to add to the scale to achieve that reliability. This option is ignored if the Compute required k items option is set to False.

 The computations for this option can only be evaluated if the reliability of the current scale is better than 0 (zero).
Creates alpha for more items Creates reliability (coefficient alpha) if a number of items are added to the scale (see the parameter Number of new items).

 The computations for this option can only be evaluated if the reliability of the current scale is better than 0 (zero).
Number of new items Specifies a number of (new) items (of the same quality) that you consider adding to the scale; STATISTICA will compute the resultant alpha based on the number of new items of the same quality. This option is ignored if the Compute alpha for more items option is set to False.

 The computations for this option can only be evaluated if the reliability of the current scale is better than 0 (zero).
Attenuation correction Creates a correlation corrected for attenuation; requires you to specify the Correlation between the current scale and the other measure, and the Reliability of the other measure.

 The attenuation correction can only be computed if the reliability of the current scale is better than 0 (zero).
Correlation with other measure Specifies the correlation between the current scale and the other measure; this option is ignored if the Compute attenuation correction option is set to False.

 The attenuation correction can only be computed if the reliability of the current scale is better than 0 (zero).
Reliability of other measure Specifies the Reliability of the other measure; this option is ignored if the Compute attenuation correction option is set to False.

 The attenuation correction can only be computed if the reliability of the current scale is better than 0 (zero).