Score an MCQ (27- or 21-item)

Usage

score_mcq(
  dat = dat,
  items = 27,
  impute_method = "none",
  round = 6,
  random = FALSE,
  trans = "none",
  return_data = FALSE,
  verbose = FALSE
)

Arguments

dat: Dataframe (longform) with subjectid, questionid, and response (0 for SIR/SS and 1 for LDR/LL)
items: Number of MCQ items: 27 (Kirby, Petry, & Bickel, 1999) or 21 (Kirby & Marakovic, 1996). Default is 27.
impute_method: One of: "none", "ggm", "GGM", "inn", "INN"
round: Numeric specifying number of decimal places (passed to base::round())
random: Boolean whether to insert a random draw (0 or 1) for NAs. Default is FALSE
trans: Transformation to apply to k values: "none", "log", or "ln". Default is "none"
return_data: Boolean whether to return the original data and new imputed responses. Default is FALSE.
verbose: Boolean whether to print subject and question ids pertaining to missing data. Default is FALSE.

Value

If return_data = FALSE (default), a summary data frame with one row per subject. If return_data = TRUE, a list with results (that summary data frame) and data (the input data plus a newresponse column reflecting any imputation).

Details

Each subject's data must satisfy a strict contract: exactly one row per canonical question id (items of them; no duplicates, no unknown ids, none missing) and responses coded 0, 1, or NA (numeric, logical, or character/factor values that coerce to 0/1). Malformed input errors rather than silently mis-scoring. Contrast with mcq_to_choice()'s lenient, ragged contract, which accepts partial per-subject coverage.

Ladder edges follow the Kaplan et al. (2014) Excel scorers. A switch point between two items scores the geometric mean of their k values, and larger-later on every item gives the first item's k. Smaller-sooner on every item follows each instrument's workbook, and the two workbooks differ: on the overall (all-item) ladder the 27-item scorer takes the geometric mean of the last item's k and the edge 0.25 (0.2494), while the 21-item scorer assigns the edge 0.1333 itself. Either way the value is a convention: those responses only show that k exceeds the steepest item's k. The small / medium / large magnitude ladders repeat their own last k (21-item: 0.1333 / 0.1292 / 0.131). The 21-item ladder keeps the published item order within each k rank, which is not strictly ascending in k; it must not be re-sorted.

Examples

score_mcq(mcq27, items = 27)
#>   subjectid overall_k  small_k medium_k  large_k geomean_k overall_consistency
#> 1         1  0.065212 0.025565 0.063690 0.064947  0.047289            0.962963
#> 2         2  0.000399 0.000633 0.001589 0.000251  0.000632            0.962963
#>   small_consistency medium_consistency large_consistency composite_consistency
#> 1                 1                  1                 1                     1
#> 2                 1                  1                 1                     1
#>   overall_proportion small_proportion medium_proportion large_proportion
#> 1           0.259259         0.333333          0.222222         0.222222
#> 2           0.777778         0.777778          0.666667         0.888889
#>   impute_method
#> 1          none
#> 2          none
dat21 <- data.frame(subjectid = 1, questionid = 1:21, response = 1)
score_mcq(dat21, items = 21)
#>   subjectid overall_k small_k medium_k large_k geomean_k overall_consistency
#> 1         1     7e-04   7e-04    7e-04   7e-04     7e-04                   1
#>   small_consistency medium_consistency large_consistency composite_consistency
#> 1                 1                  1                 1                     1
#>   overall_proportion small_proportion medium_proportion large_proportion
#> 1                  1                1                 1                1
#>   impute_method
#> 1          none