A subset of 2,000 students from the National Education Longitudinal Study of 1988 (NELS:88), focused on school discipline outcomes and their predictors. The dataset includes student demographics, family background variables, and school experience indicators related to disciplinary actions. Variables use original NELS:88 variable naming conventions (e.g., bys55a, bys12). This dataset is used for teaching logistic regression with binary outcomes and handling missing data.
Format
A tibble with 2,000 rows and 15 columns:
- bys55a
Frequency of being sent to the office. Type: numeric. Values: 0, 1, 2, 8. Where 0 = never, 1 = once or twice, 2 = more than twice, 8 = missing/not applicable. Mean = 0.47.
- bys12
Family composition. Type: numeric. Values: 1, 2, 8. Where 1 = two-parent family, 2 = other family structure, 8 = missing/not applicable.
- bys31a
Hours spent on homework per week. Type: numeric. Range: (1, 8). Ordinal scale where higher values indicate more homework hours, 8 = missing/not applicable.
- bys34a
Self-concept: English ability. Type: numeric. Range: (1, 98). Ordinal self-rating, 98 = missing/not applicable.
- bys34b
Self-concept: math ability. Type: numeric. Range: (1, 98). Ordinal self-rating, 98 = missing/not applicable.
- bys35p
Plans after high school. Type: numeric. Values: 1, 2, 3, 8. Where 1 = attend college, 8 = missing/not applicable.
- bys36c
Importance of good grades. Type: numeric. Range: (1, 8). Ordinal scale, 8 = missing/not applicable.
- sentoff
Ever sent to the office for misbehavior. Type: numeric. Binary indicator (0/1) where 1 = sent to the office at least once, 0 = never sent. NA = 28. This is the primary outcome variable for logistic regression.
- male
Sex of student. Type: numeric. Binary indicator (0/1) where 1 = male, 0 = female. NA = 20.
- race
Race/ethnicity. Type: numeric. Values: 1, 2, 3, 4. Where 1 = Asian, 2 = Hispanic, 3 = Black, 4 = White. NA = 92.
- fath_ed
Father has at least a high school education. Type: numeric. Binary indicator (0/1) where 1 = high school or above, 0 = less than high school. NA = 282.
- moth_ed
Mother has at least a high school education. Type: numeric. Binary indicator (0/1) where 1 = high school or above, 0 = less than high school. NA = 208.
- bedroom
Has own bedroom. Type: numeric. Binary indicator (0/1) where 1 = has own bedroom, 0 = does not. NA = 33.
- discuss
Frequently discusses school with parents. Type: numeric. Binary indicator (0/1) where 1 = frequently discusses, 0 = does not. NA = 33.
- osentoff
Ordinal version of sent to office. Type: numeric. Values: 0, 1, 2. Where 0 = never, 1 = once or twice, 2 = more than twice. NA = 28. Used for ordinal regression models.
Source
National Center for Education Statistics (1988). National
Education Longitudinal Study of 1988 (NELS:88). U.S. Department of
Education. Original data file: disc.dta
Details
This dataset is used in Chapters 9-13 (Logistic Regression, Multiple
Logistic Regression, Model Fit, and Ordinal Response Models) to illustrate
binary and ordinal logistic regression. Key analyses include: fitting
simple and multiple logistic regression models predicting office referrals,
computing and interpreting odds ratios, likelihood ratio tests, and
ordinal regression using osentoff as the outcome.
Note that several variables use NELS:88 coding where 8 or 98 indicates
missing/not applicable rather than NA. The derived binary and ordinal
outcome variables (sentoff, osentoff) have been recoded
with proper NA values.
Examples
data(disc)
head(disc)
#> # A tibble: 6 × 15
#> bys55a bys12 bys31a bys34a bys34b bys35p bys36c sentoff male race fath_ed
#> <int> <int> <int> <int> <int> <int> <int> <int> <int> <fct> <int>
#> 1 0 2 4 1 2 1 3 0 0 White 0
#> 2 0 1 4 2 2 1 3 0 1 White 1
#> 3 0 1 4 3 3 1 2 0 1 White 1
#> 4 0 1 4 2 2 1 3 0 1 White 1
#> 5 1 2 3 2 3 2 2 1 0 Black 1
#> 6 0 1 4 4 1 1 3 0 1 White 1
#> # ℹ 4 more variables: moth_ed <int>, bedroom <int>, discuss <int>,
#> # osentoff <int>
# Logistic regression: gender effect on office referral
glm(sentoff ~ male, data = disc, family = binomial)
#>
#> Call: glm(formula = sentoff ~ male, family = binomial, data = disc)
#>
#> Coefficients:
#> (Intercept) male
#> -1.626 1.209
#>
#> Degrees of Freedom: 1953 Total (i.e. Null); 1952 Residual
#> (46 observations deleted due to missingness)
#> Null Deviance: 2313
#> Residual Deviance: 2179 AIC: 2183
# Multiple logistic regression with demographics
glm(sentoff ~ male + factor(race) + fath_ed, data = disc, family = binomial)
#>
#> Call: glm(formula = sentoff ~ male + factor(race) + fath_ed, family = binomial,
#> data = disc)
#>
#> Coefficients:
#> (Intercept) male factor(race)Hispanic
#> -1.6235 1.2178 0.3775
#> factor(race)Black factor(race)White fath_ed
#> 1.2273 0.3771 -0.5617
#>
#> Degrees of Freedom: 1605 Total (i.e. Null); 1600 Residual
#> (394 observations deleted due to missingness)
#> Null Deviance: 1892
#> Residual Deviance: 1752 AIC: 1764