Data from an experiment comparing three methods of reading instruction on reading comprehension outcomes. Sixty-six second-grade students were randomly assigned to one of three instructional methods: Basal (traditional), DRTA (Directed Reading-Thinking Activity), and Strat (Strategies instruction). Pre-test and post-test scores were measured on multiple reading comprehension assessments. This is a primary teaching dataset used throughout the course for ANOVA, dummy variable coding, and planned comparisons.
Format
A tibble with 66 rows and 8 columns:
- subject
Student identifier. Type: numeric. Range: (1, 66).
- group
Instruction method group label. Type: character. Levels: "Basal" = standard/traditional method (reference group), "DRTA" = Directed Reading-Thinking Activity (new method), "Strat" = Strategies instruction (new method).
- pre1
Pre-test score, measure 1 (reading comprehension). Type: numeric. Range: (4, 16).
- pre2
Pre-test score, measure 2 (reading comprehension). Type: numeric. Range: (1, 13).
- post1
Post-test score, measure 1 (intruded sentences test). Type: numeric. Range: (1, 15). Primary outcome variable.
- post2
Post-test score, measure 2 (reading comprehension). Type: numeric. Range: (0, 13).
- post3
Post-test score, measure 3 (reading comprehension, standardized). Type: numeric. Range: (30, 57). Mean approximately 44.
- method
Numeric code for instruction method. Type: numeric. Values: 1 = Basal, 2 = DRTA, 3 = Strat.
Source
Schmitt, M. C. (1987). The effects of an elaborated directed
reading activity on the metacomprehension skills of third graders.
Cited in Raudenbush, S. W. & Bryk, A. S. (2002).
Hierarchical Linear Models (2nd ed.). Sage.
Original data file: reading.dta
Details
This dataset is used in Chapter 3 (Multiple Linear Regression: ANOVA) to illustrate one-way ANOVA as a regression with dummy variables, and to demonstrate planned comparisons versus post-hoc tests. Key analyses include: dummy variable coding for three-group comparisons (Basal as reference), F-test for overall group differences, planned comparisons (e.g., new methods vs. traditional, DRTA vs. Strat), Bonferroni and Scheffe adjustments for multiple comparisons, ANCOVA controlling for pre-test scores, and effect coding as an alternative to dummy coding.
Examples
data(reading)
head(reading)
#> # A tibble: 6 × 8
#> subject group pre1 pre2 post1 post2 post3 method
#> <int> <chr> <int> <int> <int> <int> <int> <int>
#> 1 1 Basal 4 3 5 4 41 1
#> 2 2 Basal 6 5 9 5 41 1
#> 3 3 Basal 9 4 5 3 43 1
#> 4 4 Basal 12 6 8 5 46 1
#> 5 5 Basal 16 5 10 9 46 1
#> 6 6 Basal 15 13 9 8 45 1
# One-way ANOVA: post-test score by instruction method
lm(post1 ~ group, data = reading)
#>
#> Call:
#> lm(formula = post1 ~ group, data = reading)
#>
#> Coefficients:
#> (Intercept) groupDRTA groupStrat
#> 6.682 3.091 1.091
#>