Data on labor force participation of 263 married Canadian women, including work status (not working, part-time, full-time), husband's income, region of residence, and presence of children. This dataset is a classic example for ordinal response modeling, as the three-level work status variable represents an ordered outcome ranging from no labor force participation to full-time employment.
Format
A tibble with 263 rows and 5 columns:
- obs
Observation number. Type: numeric. Range: (1, 263).
- husbinc
Husband's annual income in thousands of Canadian dollars. Type: numeric. Range: (1, 45). Mean approximately $14,760.
- region
Canadian region of residence. Type: character. Levels: "Atlantic" = Atlantic provinces, "Quebec" = Quebec, "Ontario" = Ontario, "Prairie" = Prairie provinces, "BC" = British Columbia.
- workstat
Work status (ordinal response variable). Type: numeric. Values: 0 = not working outside the home, 1 = working part-time, 2 = working full-time. Ordered from lowest to highest labor force participation.
- chilpres
Presence of children in the household. Type: numeric. Binary (0/1) where 1 = children present, 0 = no children present.
Source
Fox, J. (2003). Applied Regression Analysis and Generalized
Linear Models (2nd ed.). Sage. Based on data from the 1977 Canadian
Survey of Consumer Finances.
Original data file: womenlf.dta
Details
This dataset is used in Chapter 13 (Models for Ordinal Responses) to illustrate ordinal logistic regression and the proportional odds model. The three-level ordered outcome (not working, part-time, full-time) is modeled using cumulative logit links. Key analyses include: fitting the proportional odds model with husband's income, region, and children present as predictors, interpreting cumulative odds ratios, testing the proportional odds assumption, and comparing ordinal logit versus ordinal probit specifications.
Examples
data(womenlf)
head(womenlf)
#> # A tibble: 6 × 5
#> obs husbinc region workstat chilpres
#> <int> <int> <chr> <int> <int>
#> 1 1 15 Ontario 0 1
#> 2 2 13 Ontario 0 1
#> 3 3 45 Ontario 0 1
#> 4 4 23 Ontario 0 1
#> 5 5 19 Ontario 0 1
#> 6 6 7 Ontario 0 1
table(womenlf$workstat)
#>
#> 0 1 2
#> 155 42 66
# Ordinal logistic regression (requires MASS package)
# MASS::polr(factor(workstat) ~ husbinc + chilpres, data = womenlf)