Changing Categories
- collapsing categories
o recode occupational into white collar / blue collar
o categories with few subjects
o for cross tabs etc, too many categories is cumbersome
o eg, change direction & strength to strength only
- approaches to collapsing categories
o substantive approach
§ contiguous in meaning categories
o distributional approach
§ recoding variables where categories / values have natural order
· coding incomes into high / medium / low
· divide distribution into equal sized groups
o rearranging categories
§ new logic
§ eg, order industry in order of decreasing unionism
§ moving say no qualification
o reverse coding
Creating New Variables
- developing scales
- conditional transformations
- creating new variables with arithmetic transformations
o age difference (between spouses)
Standardizing Variables
- standardizing using z scores
- standardizing for different distributions
o same units, different distributions
o eg, VCE scores
- adjustments with ordinal level variables
o ranking
Problem of Missing Data
- checking for missing data bias
o group sample into 2 groups : missing / valid à how did they answer other questions
- minimizing effect of missing data
o delete cases – list wise
o delete variables
o pairwise deletion
o sample mean approach
o group means approach
o random assignment within groups
§ locate case with missing data à look at value of same variable nearest preceding case with valid code
§ introduces randomness
§ does not affect strength of correlations
o regression analysis
No comments:
Post a Comment