Posttest statistics from Lana's (1959) attitude experiment
Source:R/solomon_published_data.R
lana1959.RdThe sample size, posttest mean, and standard deviation of the four Solomon groups in one of the first experiments built on the Solomon design: a recorded pro-vivisection talk and a 10-item vivisection questionnaire given to introductory psychology classes (Lana, 1959, Table 2, p. 297). Numbers reported in the publication are reused with citation.
Format
A data frame with 4 rows, one per Solomon group, and 7 variables:
- group
The Solomon group.
- lana_group
Lana's group label (I to IV).
- pretested, treat
Indicators (1 = yes);
treatis the recorded talk.- n, mean, sd
Posttest sample size, mean, and standard deviation.
Source
Lana, R. E. (1959). Pretest-treatment interaction effects in attitudinal studies. Psychological Bulletin, 56(4), 293–300. https://doi.org/10.1037/h0044646
Details
Design. Five classes (156 students) were randomly assigned to five conditions (p. 295). Four form a Solomon design:
Group I. Pretest, then the talk 12 days later, then the posttest.
Group IV. Pretest, then the posttest 12 days later.
Group II. The talk, then the posttest.
Group III. The posttest only.
A fifth group, with a delayed posttest, is not part of the four-group design and is not included. Because whole classes were assigned, class and condition are confounded, as in elkarkri2025a.
Known result. Lana analyzed the posttest means as a two-by-two
analysis of variance (Table 3, p. 297). The treatment effect was
significant, F = 5.35, and the pretest and the interaction were not.
solomon_from_summary() reproduces it within rounding (treatment
F = 5.36 on 1 and 152 df). Lana's sums of squares are on the scale of the
cell means (an unweighted-means analysis). Dividing the package's sums of
squares by the harmonic mean of the cell sizes recovers them, and the F
tests are the same.
Historical note. Lana concluded that the pretest did not sensitize participants to the talk (p. 298), the opposite of Solomon's (1949) spelling result. Together they begin the record that Willson and Putnam (1982) later pooled.
Examples
lana1959
#> group lana_group pretested treat n mean sd
#> 1 Pretested, treatment I 1 1 26 42.96 9.06
#> 2 Pretested, control IV 1 0 32 40.28 5.48
#> 3 Unpretested, treatment II 0 1 50 42.90 5.34
#> 4 Unpretested, control III 0 0 48 40.77 5.76
with(lana1959, solomon_from_summary(n, mean, sd))
#> Solomon analysis from summary statistics
#> ----------------------------------------
#> Pooled error variance: 39.077 on 152 df (equal variances assumed)
#>
#> Two-way ANOVA on the posttest (Type III sums of squares)
#> Treatment SS = 209.291 df = 1 F = 5.36 p = 0.022
#> Pretest SS = 1.673 df = 1 F = 0.04 p = 0.836
#> Treatment x Pretest SS = 2.736 df = 1 F = 0.07 p = 0.792
#> Error SS = 5939.644 df = 152
#>
#> Contrasts with 95% confidence intervals
#> Test A: Pretest x Treatment 0.550 [-3.556, 4.656], t(152) = 0.26, p = 0.792
#> Test B: Treatment | pretested 2.680 [-0.581, 5.941], t(152) = 1.62, p = 0.106
#> Test C: Treatment | unpretested 2.130 [-0.366, 4.626], t(152) = 1.69, p = 0.094
#> Test D: ATE (avg over pretest) 2.405 [0.352, 4.458], t(152) = 2.31, p = 0.022
#> Pretest main effect -0.215 [-2.268, 1.838], t(152) = -0.21, p = 0.836