• Merge raw data from multiple sources with minimal pre-processing;
• Check whether item responses are comparable across sources;
• Clean data to establish item comparability:
○ Ensure constant directionality/polarity:
■ Review content and response options;
■ Run correlation matrices, flag items with sizable negative correlations;
■ Reverse code as necessary.
○ Ensure consistency in scoring type and scales:
■ Review response options;
■ Cross-tabulate items across datasets to evaluate whether items have different minimum and maximum values by dataset;
■ Exclude summary scores and counts in favor of more granular data;
■ Truncate, collapse response categories as necessary.
○ Eliminate conditional dependency:
■ Review content and logic flows;
■ Perform parametric modeling, scrutinize output for residuals;
■ Exclude conditional items.
○ Address missingness/skewness:
■Tabulate frequency of each item being endorsed;
■ Filter out items with coded missingness;
■ Filter out items with same min and max within a dataset;
■ Truncate, collapse response categories as necessary;
■Exclude items with no variability.
• Establish configural invariance:
○ Estimate parametric models within each dataset;
○ Scrutinize output for residuals;
○ Include residual covariances for items having high covariance residuals.