CODES 2000 User Forum -- Data Network Note #16
Step by Step Instructions for Linkage, Imputation, and Analysis
Applies to: CODES Data Network.
Last updated: Friday May 17, 2002.
SUMMARY
The methodology that you must follow in order to reply to a CODES data request is fairly complicated. Here, we present step by step instructions as an outline with references to earlier, more detailed information.
METHODOLOGY
- Install CODES 2000 Release 2.2.373 or later.
In a few states, our earlier linkage and imputation methodology resulted in thousands of matched record pairs being assigned to the same set. Consequently, most of these matches were not included in the final one-to-one linkage, substantially reducing the apparent coverage. Other states may have experienced less severe versions of this problem that limited their coverage. The new software release provides a better methodology that should avoid the problem by assigning record pairs to sets after imputation, rather than before.
See Tech Note # 14 -- Installation Notes for CODES 2000 Release 2.2.373
- Plan your linkage, imputation, and analysis strategies.
Review documentation about the data files.
Confirm unique record identifiers.
Plan how to select cases. See Tech Note # 9 - Case Selection Versus Case Classification.
Plan how to estimate total links.
Identify fields that might be useful for matching.
Identify ancillary data that might be useful for matching. See Help Topic - View Reference Locations.
Plan standardization methods. See Help Topic - View Standardization Methods.
Plan custom Visual Basic functions. See Tech Note # 12 - Using Custom Functions in Queries.
Plan your match strategy. See Tech Note # 8 - Choosing a Match Strategy. See Data Network Note # 13 - Finding Most of your Estimated Total Links.
Plan comparison methods. See Help Topic - View Comparison Methods.
See Tech Note # 16 -- Imputing Complete, One-to-One Dual Linkages
Plan value imputation model. See SAS on-line documentation for PROC MI.
Plan regression model or other statistical analysis method. Your analysis method must be supported by SAS PROC MIANALYZE. See SAS on-line documentation for PROC MIANALYZE.
- Create and open a new CODES 2000 project to begin implementing your linkage strategy.
See Help Topic - View CODES Projects.
You can import an existing project into your new project. You can import specifications only, specifications plus data, or specifications plus data plus match results.
See Tech Note # 15 -- Importing an Existing Project into a New Project
- Prepare the Crash data source.
4.1 Select Input Data for Crash. See Help Topic - Select Input Data.
4.2 Import Data for Crash. See Help Topic - Import Selected Data.
4.3 Analyze In Fields for Crash. See Help Topic - Analyze Input Fields.
4.4 Standardize Input for Crash. See Help Topic - Standardize Input.
4.5 Analyze Standard Fields for Crash. See Help Topic - Analyze Standard Fields.
4.6 Compact Database for Crash. See Help Topic - Compact the Database.
- Prepare the Hospital data source.
5.1 Select Input Data for Hospital. See Help Topic - Select Input Data.
5.2 Import Data for Hospital. See Help Topic - Import Selected Data.
5.3 Analyze In Fields for Hospital. See Help Topic - Analyze Input Fields.
5.4 Standardize Input for Hospital. See Help Topic - Standardize Input.
5.5 Analyze Standard Fields for Hospital. See Help Topic - Analyze Standard Fields.
5.6 Compact Database for Hospital. See Help Topic - Compact the Database.
- Prepare and conduct the Crash to Hospital Dual Match with a 0.001 cutoff probability.
6.1 Prepare Specs. See Help Topic - Prepare Link Specifications. You do not have to include both person and event variables as join fields in every pass.
6.2 Perform All Match Passes. See Help Topic - Perform a Match Pass.
6.3 Review All Match Passes. See Help Topic - Review Match Results.
6.4 Impute Complete One to One Linkages. See Tech Note # 16 - Imputing Complete, One to One Dual Linkages.
6.5 Review your linkage results and check the fit of your linkage probability model. Correct any field error rates or other model parameters that appear to be inaccurate.
- Conduct missing value Multiple Imputation for your linked dataset.
See SAS on-line documentation for PROC MI for a discussion of multiple imputation theory and practice.
See Data Network Note # 14 - Procedures for Multiple Imputation (Phase II).
- Conduct regression analysis of your imputed datasets.
See SAS Help Topic for PROC REG or PROC LOGISTIC for a discussion of regression analysis theory and practice.
See Data Network Note # 14 - Procedures for Multiple Imputation (Phase II).
- Combine regression results into final parameter estimates.
See SAS on-line documentation for PROC MIANALYZE for a discussion of multiple imputation theory and practice.
See Data Network Note # 14 - Procedures for Multiple Imputation (Phase II).
- Conduct sensitivity analyses.
See Data Network Note # 8 - Variations in Multiple Imputation Results.
See Data Network Note # 14 - Procedures for Multiple Imputation (Phase II).