How do you determine correlation (corrected for overlap) in psychometric correlation analysis in SPSS?
In correlation, when researcher consider the bivariate correlation, also called Zero Order Correlation, the full strength of correlation between the 2 variables will be resulted.
However, there are Partial and semi partial correlation, in which the overlap is avoided. In Statistical programs such as SPSS, you can find this option under correlation and regression. Actually in multiple linear regression, the statistical program consider this automatically; by giving the highest correlation the first order, then remove the overlap from the 2nd strongest correlated variable..etc.
I am not sure what you are looking for, but if you want to see the correlation of the item with scale if the item is deleted (Corrected Item-Total Correlation), you can use:
RELIABILITY
/VARIABLES=var01 to varn
/SCALE('test scale 01') ALL
/MODEL=ALPHA
/STATISTICS=DESCRIPTIVE SCALE
/SUMMARY=TOTAL.
IF you only use drop-down-menu's, go to scale -> reliability analysis -> statistics and mark 'scale if item deleted'.
If you want to calculate the correlation between two scales corrected for error, you have to do it by hand or in excel using the formula given in McNunnally. SPSS has a procedure, but only when you have two similar scales that are viewed as split-half or equivalent. Calculating by hand or in excel is much faster.
In correlation, when researcher consider the bivariate correlation, also called Zero Order Correlation, the full strength of correlation between the 2 variables will be resulted.
However, there are Partial and semi partial correlation, in which the overlap is avoided. In Statistical programs such as SPSS, you can find this option under correlation and regression. Actually in multiple linear regression, the statistical program consider this automatically; by giving the highest correlation the first order, then remove the overlap from the 2nd strongest correlated variable..etc.
The biserial correlation between an item and the total test of which the item is a part tends to be misleadingly high when used in item analysis, since the item is included in the total test. Two formulas with correction for this overlap are derived and compared with Zubin's and Guilford's formulas. One of the new coefficients is invariant to test length.
Article Correction of Item-Total Correlations in Item Analysis
- Badges
- Science topic
- Similar topics
- Psychology
- Psychometrics