Looks up scored PID-5 columns in the published normative tables shipped as
pid_norms and returns a T score and a percentile for each. Scores are
produced by score_pid5() and validity_pid5(); this function converts them
and never rescores.
Arguments
- data
A data frame containing scored PID-5 columns.
- scores
The score columns to convert, as column names or column positions (mirroring the
itemsargument ofscore_pid5()). Each column must be numeric (or logical) and each may be named only once.- version
Which PID-5 version the scores came from:
"FULL"(220 items),"SF"(100 items), or"BF"(25 items). The normative tables differ by version.- srange
The response range the items were coded on, as
c(low, high). Any four-option coding is accepted and reconciled to the officialc(0, 3)range before lookup; a coding with a different number of options is not convertible. See Details.- prefix
The prefix
score_pid5()applied to its output columns, used to match a score column back to its scale. Matched literally, not as a regular expression: a column name that does not begin with exactly this string keeps its whole name and is reported as uncovered. Pass""when the columns are named for the scales themselves, with no prefix to strip.- append
Whether to return the input
datawith the conversion columns appended (TRUE, the default) or the conversion columns alone.
Value
A tibble with a _t column for every converted scale
whose normative rows carry a T score (the five domains, the 25 facets of
the full and short forms, and the brief form's total) and a _ptl column
for every converted scale, alongside all original data columns if
requested. The four validity scales (INC, INCS, ORS, PRD) are
distributed as percentiles only and get no _t column.
Details
For each named score column the tables are searched for the printed row whose raw score is nearest the observed value, and that one row's printed T score and printed percentile are returned. Nothing is interpolated: every returned number is a cell of Markon et al. (2024).
Markon et al. print the tables but give no instruction for reading them — no rounding, interpolation, or tie rule appears anywhere in the book — so the rules below are this package's, chosen and defended rather than quoted. They were settled by an independent review of the tables recorded in the package's development history as report RR02 (in the project repository, not the installed package).
Between printed rows. The nearer row wins. Printed raws step by 0.01-0.07 while attainable scores fall on much coarser grids (a 5-item brief-form domain mean can only be a multiple of 0.2), so most lookups land between rows.
Ties. Where two or more rows are equally near – a raw printed in several rows, or a value exactly midway between two rows – the row whose T score is nearest 50 is returned. The four validity scales carry no T score, so a tie there returns the row whose percentile is nearest 0.50.
Scores of 0. Nearly every tabled scale prints raw 0.00 across a run of low T scores, because the linear T the book tabulated predicts a negative raw there and 0.00 is printed instead. The tie rule returns the run's highest T, the one row of the run that renders an attainable score. (Two columns – full-form detachment and risk taking – print 0.00 just once, so no tie arises there.) Its printed percentile is positive on some scales and 0.00 on others; that asymmetry is a property of the published tables, not of this function.
Scores outside the table. A score above the highest printed raw returns whatever an observation at that raw returns, rather than an extrapolation – the last printed row on most scales, but the lowest-T row of the run where the top raw is printed several times (see "Unattainable printed rows" below). A score below the lowest returns whatever an observation at the lowest printed raw returns – which, on the scales whose tables print a run of 0.00, is that run's highest-T row and not the table's first row, so the two agree instead of jumping. A warning reports how many observations were capped at each end. This is reachable in ordinary data:
PRDis a 22-item sum reaching 66 while its table stops at 55.Unattainable printed rows. 47 of the 66 tabled scales print rows above the 3.00 ceiling a 0-3 item mean can reach – five domain scales and 42 of the 50 facet columns – so the top of those T ranges cannot be attained. A maximum score returns T = 84 (brief-form negative affectivity), 87 (brief-form detachment), 93 (brief-form disinhibition), 87 (full-form negative affectivity), 85 (short-form negative affectivity), or, on a facet, a T between 76 and 98 depending on the facet and the form – each at percentile 1.00. Nineteen facet columns go further and print the same top raw, 4.00, on several consecutive T rows (12 of them for short-form anxiousness); those rows ship as published, and the tie rule above returns the lowest T of the run. Nothing is wrong with such data and nothing is reported.
Comparison tolerance. All comparisons use an absolute tolerance of 1e-8, so that scores on grids with no exact binary representation (a short-form domain mean is a twelfth) match the printed 2-decimal raws as intended.
Columns the tables do not cover for the requested version – SD-TD, or
any facet on the brief form, for instance – return NA in both conversion
columns with a warning naming them. An NA score returns NA.
Reporting and silence. Everything this function reports – the capping
count above, the uncovered-column warning, and the two response-coding
reports below – is a warning condition, so a single suppressWarnings()
call silences the function and any one report can still be caught and
tested for individually.
Errors. scores is checked before anything is converted. Naming the
same score column twice is an error rather than a duplicated pair of output
columns, and a factor or character score column is an error rather than a
silent coercion – a factor's integer codes are not its scores, and a
character column would coerce to NA. Logical columns are accepted, since
a 0/1 indicator converts as it reads.
Response coding. The normative tables are built on the official
four-option 0-3 coding. Data collected on a four-option coding that merely
starts elsewhere – 1-4, say – carries the same information, so each score
is reconciled to the official range before lookup and the conversion
proceeds. A coding with a different number of options is a different
metric: no mapping onto a four-option norm table is defined, so every
conversion column is returned as NA with a warning, and the items must be
recoded and rescored.
How much a shift moves a score depends on how the scale is computed, so the reconciliation is applied per scale rather than per item. No published source states these rules – Markon et al. give the tables for the official coding only – so, like the reading rules above, they are this package's, derived from each scale's own definition:
Item means (the 25 facets, the five domains, and the brief form's total) are reconciled by subtracting
srange[[1]]. Shifting every item by a constant shifts their mean by the same constant.PRDis a plain sum over its 22 items, so the same shift moves it bysrange[[1]]times the number of items, which is what is subtracted. The item count is read from pid_items rather than assumed.INCandINC-Sare sums of absolute differences within item pairs. A constant added to both members of a pair cancels in the difference, so these are unchanged by a shift and nothing is subtracted.ORSis a count of items answered at the top of the response range –validity_pid5()computes it by comparing each item tosrange[[2]]rather than to a fixed value (R/validity_pid5.R:153in the package sources). A shift moves the top of the range along with the answers, so the same items are counted and the score is unchanged.
A shifted coding raises one warning per call naming which of the requested scales were adjusted and which were left alone; where every requested scale turns out to be coding-invariant, it says so rather than claiming an adjustment. The warning covers the scales the tables actually carry, so a request the tables cover nowhere raises the coverage warning above instead and nothing about coding. The official coding is silent.
One consequence is worth stating plainly, because it can put two differently
grounded numbers side by side in the same session: validity_pid5()'s
published cut scores are not reconciled to a shifted coding. PRD and
SD-TD are compared against fixed thresholds that assume 0-3 items, and
validity_pid5() warns rather than adapting them. So a respondent scored on
a 1-4 coding can receive a reconciled percentile from this function and, from
validity_pid5(), a validity flag still read against the 0-3 thresholds.
Adapting those cut scores is a separate, deliberately deferred question.
References
Markon, K. E., Fossati, A., Somma, A., & Krueger, R. F. (2024). Understanding the Personality Inventory for DSM-5 (PID-5). American Psychiatric Association Publishing. The normative tables in pid_norms, Appendix "Normative Score Distributions" (pp. 113-219), are the source for every value this function returns.
Examples
# Score the brief form, then convert its domains and total
scored <- score_pid5(sim_pid5bf, items = 1:25, version = "BF")
norm_pid5(scored, scores = paste0("pid_", c("detachment", "total")),
version = "BF", append = FALSE)
#> # A tibble: 100 × 4
#> pid_detachment_t pid_detachment_ptl pid_total_t pid_total_ptl
#> <int> <dbl> <int> <dbl>
#> 1 65 0.92 76 0.98
#> 2 74 0.98 77 0.98
#> 3 58 0.77 78 0.99
#> 4 74 0.98 71 0.95
#> 5 58 0.77 80 0.99
#> 6 49 0.59 69 0.94
#> 7 71 0.96 69 0.94
#> 8 68 0.96 72 0.95
#> 9 52 0.59 72 0.95
#> 10 68 0.96 66 0.93
#> # ℹ 90 more rows