umx_score_scale: Score a psychometric scale by summing normal and reversed...

umx_score_scaleR Documentation

Score a psychometric scale by summing normal and reversed items.

Description

Use this function to generate scores as the appropriate sum of responses to the normal and reversed items in a scale.

Items must be named on the pattern basename + N + suffix, where base is the prefix common to all item (column) names, N is item number in the scale, and suffix an optional trail (like "_T1").

pos and rev are vectors of the item numbers for the normal and reverse-scored item numbers.

To reverse items, the function uses max and min as the lowest and highest possible response scores to compute how to reverse items.

note: min defaults to 1.

Usage

umx_score_scale(
  base = NULL,
  pos = NULL,
  rev = NULL,
  min = 1,
  max = NULL,
  data = NULL,
  score = c("total", "proportionCorrect", "errors", "mean", "max", "factor"),
  name = NULL,
  na.rm = TRUE,
  minManifests = NA,
  alpha = FALSE,
  mapStrings = NULL,
  correctAnswer = NULL,
  omegaNfactors = 1,
  digits = 2,
  verbose = FALSE,
  suffix = ""
)

Arguments

base

String common to all item names.

pos

The positive-scored item numbers.

rev

The reverse-scored item numbers.

min

Minimum legal response value (default = 1). Not implemented for values other than 1 so far...

max

Maximum legal response value (also used to compute reversed item values).

data

The data frame

score

Score total (default), proportionCorrect, errors, mean, max, or factor scores

name

The name of the scale to be returned. Defaults to "base_score"

na.rm

Whether to delete NAs when computing scores (Default = TRUE) Note: Choice affects mean!

minManifests

How many missing items to tolerate for an individual (when score = factor)

alpha

print Reliability (omega and Cronbach's alpha) (TRUE)

mapStrings

Recoding input like "No"/"Maybe"/"Yes" into numeric values (0,1,2)

correctAnswer

Use when scoring items with one correct response (1/0).

omegaNfactors

Number of factors for the omega reliability (default = 1)

digits

Rounding for omega etc. (default 2)

verbose

Whether to print the whole omega output (FALSE)

suffix

(if dealing with, e.g. "_T1")

Details

In the presence of NAs, score= "mean" and score = "totals" both return NA unless na.rm = TRUE. score = "max", ignores NAs no matter what.

Value

  • scores

References

  • McNeish, D. (2018). Thanks coefficient alpha, we’ll take it from here. Psychological Methods, 23, 412-433. \Sexpr[results=rd]{tools:::Rd_expr_doi("10.1037/met0000144")}.

See Also

Other Data Functions: noNAs(), prolific_anonymize(), prolific_check_ID(), prolific_read_demog(), umxFactor(), umxHetCor(), umx_as_numeric(), umx_cont_2_quantiles(), umx_lower2full(), umx_make_MR_data(), umx_make_TwinData(), umx_make_fake_data(), umx_make_raw_from_cov(), umx_merge_randomized_columns(), umx_polychoric(), umx_polypairwise(), umx_polytriowise(), umx_read_lower(), umx_rename(), umx_reorder(), umx_select_valid(), umx_stack(), umx_strings2numeric(), umx

Examples

library(psych)
library(psychTools)
data(bfi)

# ==============================
# = Score Agreeableness totals =
# ==============================

# Handscore subject 1
# A1(R)+A2+A3+A4+A5 = (6+1)-2 +4+3+4+4 = 20

tmp = umx_score_scale(base = "A", pos = 2:5, rev = 1, max = 6, data= bfi, name = "A")
tmp[1, namez(tmp, "A",ignore.case = FALSE)]
#  A1 A2 A3 A4 A5  A
#  2  4  3  4  4  20

# ====================
# = Request the mean =
# ====================
tmp = umx_score_scale(name = "A", base = "A", 
   pos = 2:5, rev = 1, max = 6, data= bfi, score="mean")
tmp$A[1] # = 4

# ========================
# = Request factor score =
# ========================
## Not run: 
tmp = umx_score_scale(name = "A", base = "A", pos = 2:5, rev = 1,
   max = 6, score = "factor", minManifests = 4, data= bfi)
#            g
# A2 0.6574826
# A3 0.7581274
# A4 0.4814788
# A5 0.6272332
# A1 0.3736021

# ==================
# = Request alpha  =
# ==================

tmp=umx_score_scale(base="A", pos=2:5, rev=1, max=6, data=bfi, alpha=TRUE)
# omega t = 0.72

## End(Not run)

# ==================
# = na.rm = TRUE ! =
# ==================
tmpDF = bfi
tmpDF[1, "A1"] = NA
tmp = umx_score_scale("A", pos = 2:5, rev = 1, max = 6, data= tmpDF, score="mean")
tmp$A_score[1] # 3.75

tmp= umx_score_scale("A", pos= 2:5, rev= 1, max = 6, data = tmpDF,
   score="mean", na.rm=FALSE)
tmp$A_score[1] # NA (reject cases with missing items)

# ===============
# = Score = max =
# ===============
tmp = umx_score_scale("A", pos = 2:5, rev = 1, max = 6,
  data = bfi, name = "A", score = "max")
tmp$A[1] # Subject 1 max = 5 (reversed) item 1

# Default scale name
tmp = umx_score_scale("E", pos = 3:5, rev = 1:2, max = 6, 
   data= tmp, score = "mean", na.rm = FALSE)
tmp$E_score[1]

# Using @BillRevelle's psych package: More diagnostics, including alpha
scores= psych::scoreItems(items = bfi, min = 1, max = 6, keys = list(
	E = c("-E1","-E2", "E3", "E4", "E5"),
	A = c("-A1", "A2", "A3", "A4", "A5")
))
summary(scores)
scores$scores[1, ]
#  E   A 
# 3.8 4.0 

# Compare output
# (note, by default psych::scoreItems replaces NAs with the sample median...)
RevelleE = as.numeric(scores$scores[,"E"])
RevelleE == tmp[,"E_score"]

# =======================
# = MapStrings examples =
# =======================
mapStrings = c(
   "Very Inaccurate", "Moderately Inaccurate", 
   "Slightly Inaccurate", "Slightly Accurate",
   "Moderately Accurate", "Very Accurate")
bfi$As1 = factor(bfi$A1, levels = 1:6, labels = mapStrings)
bfi$As2 = factor(bfi$A2, levels = 1:6, labels = mapStrings)
bfi$As3 = factor(bfi$A3, levels = 1:6, labels = mapStrings)
bfi$As4 = factor(bfi$A4, levels = 1:6, labels = mapStrings)
bfi$As5 = factor(bfi$A5, levels = 1:6, labels = mapStrings)
bfi= umx_score_scale(name="A" , base="A", pos=2:5, rev=1, max=6, data=bfi)
bfi= umx_score_scale(name="As", base="As", pos=2:5, rev=1, mapStrings = mapStrings, data= bfi)

tbates/umx documentation built on March 16, 2024, 4:26 a.m.