Statistical data coding - Category Home
Statistical data coding - Topics
Duplicate records from a double submission and the fuzzy match that finds them
Building the data dictionary while collecting instead of after the spreadsheet exists
Binning a continuous variable into quartiles throws away power for no reason
Masking the group labels from the analyst so the statistics are done blind
Missing values coded as -99 that leak into a mean if never set to system missing
A codebook without units and labels makes a merged file unreadable
Date formats and time zones that shift a record by a day after the merge
Free-text fields full of typos before any coding can start
Reverse-scoring negatively worded items before the scale total is computed
A de-identification mapping table kept in a separate encrypted file
Don't know vs refused vs left blank in a survey dataset
Overdispersion in a count model and standard errors that look too small
Collapsing a sparse category with two cases into an other group
Splitting a continuous measure into high and low groups throws away the information
A composite score hides the two items that were doing all the work
Dummy coding a categorical variable and the reference category nobody documents
Versioning a codebook when a variable is recoded after the first analysis
Multiple imputation vs complete-case analysis: two answers from one dataset