Score items using regression or correlation based weights
Item weights from bestScales or lmCor are used to find weighted scale scores. In contrast to the unit weights used in scoreItems, scoreWtd will multiply the data by a set of weights to find scale scores. These weight may come from a regression (e.g., lm or lmCor) or may be the zero order correlation weights from bestScales.
scoreWtd(weights, items, std = TRUE, sums = FALSE, impute = "none")
weights: This is just a matrix of weights to use for each item for each scale.items: Matrix or dataframe of raw item scoresstd: if TRUE, then find weighted standard scores else just use raw datasums: By default, find the average item score. If sums = TRUE, then find the sum scores. This is useful for regression with an intercept termimpute: impute="median" replaces missing values with the item medians, impute = "mean" replaces values with the mean response. impute="none" the subject's scores are based upon the average of the keyed, but non missing scores. impute = "none" is probably more appropriate for a large number of missing cases (e.g., SAPA data).Although meant for finding correlation weighted scores using the weights from bestScales, it also possible to use alternative weight matrices, such as those returned by the coefficients in lm.
A data frame of scores.
William Revelle
bestScales and lmCor
#find the weights from a regression model and then apply them to a new set #derivation of weights from the first 20 cases model.lm <- lm(rating ~ complaints + privileges + learning,data=attitude[1:20,]) #or use lmCor to find the coefficents model <- lmCor(rating ~ complaints + privileges +learning,data=attitude[1:20,],std=FALSE) #Apply these to a different set of data (the last 10 cases) #note that the regression coefficients need to be a matrix scores.lm <- scoreWtd(as.matrix(model.lm$coefficients),attitude[21:30,],sums=TRUE,std=FALSE) scores <- scoreWtd(model$coefficients,attitude[21:30,],sums=TRUE,std=FALSE) describe(scores)