Validate the structure and coding of a Solomon four-group design

Checks that data can support a Solomon four-group analysis before a model is fitted: equal input lengths, 0/1 coding of the design indicators, all four cells present, enough observed outcomes per cell, and the distinction between structurally absent and incidentally missing values (see check_solomon_missing()). Problems are returned as a table of issues rather than stopping at the first one, so every problem is reported at once.

Usage

validate_solomon(y_post, treat, pretested, y_pre = NULL, min_cell_n = 2)

Arguments

y_post: Numeric posttest scores.
treat: Treatment indicator coded 0/1 (or logical).
pretested: Pretest indicator coded 0/1 (or logical).
y_pre: Optional numeric pretest scores, missing by design for unpretested participants.
min_cell_n: Minimum number of observed posttest scores required in each cell (at least 2).

Value

An object of class solomon_validation with valid (TRUE when no errors were found), issues (severity, check, and message), cells (counts by cell), and missing (the check_solomon_missing() result).

Details

Accepted coding: treat and pretested must be numeric 0/1 or logical. Factors and character codes are rejected so that group membership is never inferred from level order. The Solomon design requires all four cells (Solomon, 1949): pretested treatment, pretested control, unpretested treatment, and unpretested control.

Severity:

error: the data cannot support a Solomon analysis as supplied (unequal lengths, invalid coding, an empty cell, fewer than min_cell_n observed posttest scores in a cell, or no observed pretests among pretested participants when y_pre is supplied).
warning: analyses can run, but some participants are excluded or values are inconsistent with the design (unassigned participants, incidental missingness, pretest scores recorded for unpretested participants).
note: information for reporting, such as the range of cell sizes.

min_cell_n is a technical minimum, not a sample-size recommendation: at least two observations per cell are needed to estimate within-cell variability, and HC3 standard errors can be undefined for a cell with a single observation. Plan cell sizes with a power analysis.

References

Solomon, R. L. (1949). An extension of control group design. Psychological Bulletin, 46(2), 137-150.

Examples

data(solomon_example)

# A valid design.
with(solomon_example, validate_solomon(y_post, treat, pretested, y_pre))
#> Solomon design validation: no errors found
#> 
#>  Group                   Cell  n Post missing Pre absent (design) Pre missing
#>      1   Pretested, treatment 30            0                   0           0
#>      2     Pretested, control 30            0                   0           0
#>      3 Unpretested, treatment 30            0                  30           0
#>      4   Unpretested, control 30            0                  30           0
#>  Pre unexpected
#>               0
#>               0
#>               0
#>               0
#> 
#> [NOTE] Cell sizes range from 30 to 30.

# An empty cell makes the design invalid.
no_group_3 <- solomon_example[!(solomon_example$pretested == 0 & solomon_example$treat == 1), ]
with(no_group_3, validate_solomon(y_post, treat, pretested, y_pre))
#> Solomon design validation: errors found
#> 
#>  Group                   Cell  n Post missing Pre absent (design) Pre missing
#>      1   Pretested, treatment 30            0                   0           0
#>      2     Pretested, control 30            0                   0           0
#>      3 Unpretested, treatment  0            0                   0           0
#>      4   Unpretested, control 30            0                  30           0
#>  Pre unexpected
#>               0
#>               0
#>               0
#>               0
#> 
#> [ERROR] The Solomon design requires all four cells; no participants are in:
#>   Unpretested, treatment.