item.scores: Scale Scores

Description

This function computes (prorated) scale scores by averaging the (available) items that measure a single construct by default.

Usage

item.scores(..., data = NULL, fun = c("mean", "sum", "median", "var", "sd", "min", "max"),
            prorated = TRUE, p.avail = NULL, n.avail = NULL, append = TRUE,
            name = "scores", as.na = NULL, check = TRUE)

Value

Returns a numeric vector with the same length as nrow(x) containing (prorated) scale scores.

Arguments

...: a matrix or data frame with numeric vectors. Alternatively, an expression indicating the variable names in data e.g., item.scores(x1, x2, x3, data = dat). Note that the operators ., +, -, ~, :, ::, and ! can also be used to select variables, see 'Details' in the df.subset function.
data: a data frame when specifying one or more variables in the argument .... Note that the argument is NULL when specifying a matrix or data frame for the argument ....
fun: a character string indicating the function used to compute scale scores, default: "mean".
prorated: logical: if TRUE (default), prorated scale scores are computed (see 'Details'); if FALSE, scale scores of only complete cases are computed.
p.avail: a numeric value indicating the minimum proportion of available item responses needed for computing a prorated scale score for each case, e.g. p.avail = 0.8 indicates that scale scores are only computed for cases with at least 80% of item responses available. By default prorated scale scores are computed for all cases with at least one item response. Note that either argument p.avail or n.avail is used to specify the proration criterion.
n.avail: an integer indicating the minimum number of available item responses needed for computing a prorated scale score for each case, e.g. n.avail = 2 indicates that scale scores are only computed for cases with item responses on at least 2 items. By default prorated scale scores are computed for all cases with at least one item response. Note that either argument p.avail or n.avail is used to specify the proration criterion.
append: logical: if TRUE (default), a variable with scale scores is appended to the data frame specified in the argument data.
name: a character string indicating the names of the variable appended to the data frame specified in the argument data when append = TRUE. By default, the variable is named scores.
as.na: a numeric vector indicating user-defined missing values, i.e. these values are converted to NA before conducting the analysis.
check: logical: if TRUE (default), argument specification is checked.

Author

Takuya Yanagida takuya.yanagida@univie.ac.at

Details

Prorated mean scale scores are computed by averaging the available items, e.g., if a participant answers 4 out of 8 items, the prorated scale score is the average of the 4 responses. Averaging the available items is equivalent to substituting the mean of a participant's own observed items for each of the participant's missing items, i.e., person mean imputation (Mazza, Enders & Ruehlman, 2015) or ipsative mean imputation (Schafer & Graham, 2002).

Proration may be reasonable when (1) a relatively high proportion of the items (e.g., 0.8) and never fewer than half are used to form the scale score, (2) means of the items comprising a scale are similar and (3) the item-total correlations are similar (Enders, 2010; Graham, 2009; Graham, 2012). Results of simulation studies indicate that proration is prone to substantial bias when either the item means or the inter-item correlation vary (Lee, Bartholow, McCarthy, Pederson & Sher, 2014; Mazza et al., 2015).

References

Enders, C. K. (2010). Applied missing data analysis. New York, NY: Guilford Press.

Graham, J. W. (2009). Missing data analysis: Making it work in the real world. Annual Review of Psychology, 60, 549-576. https://doi.org/10.1146/annurev.psych.58.110405.085530

Graham, J. W. (2012). Missing data: Analysis and design. New York, NY: Springer

Lee, M. R., Bartholow, B. D., McCarhy, D. M., Pederson, S. L., & Sher, K. J. (2014). Two alternative approaches to conventional person-mean imputation scoring of the self-rating of the effects of alcohol scale (SRE). Psychology of Addictive Behaviors, 29, 231-236. https://doi.org/10.1037/adb0000015

Mazza, G. L., Enders, C. G., & Ruehlman, L. S. (2015). Addressing item-level missing data: A comparison of proration and full information maximum likelihood estimation. Multivariate Behavioral Research, 50, 504-519. https://doi.org/10.1080/00273171.2015.1068157

Schafer, J. L., & Graham, J. W. (2002). Missing data: Our view of the state of the art. Psychological Methods, 7, 147-177.' https://doi.org/10.1037/1082-989X.7.2.147

Examples

Run this code

dat <- data.frame(item1 = c(3,  2,  4, 1,  5, 1,  3, NA),
                  item2 = c(2,  2, NA, 2,  4, 2, NA,  1),
                  item3 = c(1,  1,  2, 2,  4, 3, NA, NA),
                  item4 = c(4,  2,  4, 4, NA, 2, NA, NA),
                  item5 = c(3, NA, NA, 2,  4, 3, NA,  3))

# Example 1a: Prorated mean scale scores
item.scores(dat)

# Example 1b: Alternative specification using the 'data' argument
item.scores(., data = dat)

# Example 2: Prorated standard deviation scale scores
item.scores(dat, fun = "sd")

# Example 3: Sum scale scores without proration
item.scores(dat, fun = "sum", prorated = FALSE)

# Example 4: Prorated mean scale scores,
# minimum proportion of available item responses = 0.8
item.scores(dat, p.avail = 0.8)

# Example 5: Prorated mean scale scores,
# minimum number of available item responses = 3
item.scores(dat, n.avail = 3)

Run the code above in your browser using DataLab