Detect SCD design type — detect_design

Phase sequences are built per tier (see tier_vars()) from time-ordered run lengths (rle), so repeated phases such as A-B-A-B are preserved rather than collapsed to A-B, and a case identifier repeated across studies is read as separate series.

Usage

detect_design_type(data)

Arguments

data: A data frame with case and phase columns.

Value

A character string: "ab", "reversal", "multiple_baseline", "alternating", "changing_criterion", or "unknown".

Details

Checks run in this order: alternating treatments (two or more conditions alternating in short runs after an optional baseline phase), changing criterion with a numeric criterion column that changes two or more times, up or down, between adjacent intervention sessions of a tier (so A-B-B-B-A-B with stepped criteria is not read as a reversal), reversal (three or more runs alternating between baseline-role and intervention-role phases, with Withdrawal, Return, Reversal, or a numbered baseline in the baseline role and Reinstatement in the intervention role), sequential non-recurring designs (A-B, A-B-M, A-B1-B2-M, A-B-M-G with onsets that differ across tiers), and changing criterion (four or more conditions after the first whose labels carry the intervention role or no known role, such as B1/B2/B3, possibly returning to an earlier criterion). Maintenance or generalization phases rule out changing criterion: a single A-B-M-G series is "ab".

Examples

d <- data.frame(
  case = rep(c("C1", "C2"), each = 10),
  phase = rep(c("A", "B"), 10),
  time = rep(1:10, 2),
  outcome = rnorm(20)
)
detect_design_type(d)
#> [1] "alternating"