quest.reliability {concord} | R Documentation |
Calculates a few common measures of questionnaire reliability
quest.reliability(x,nsplits=10)
x |
a matrix of numeric scores with subjects as rows and items as columns |
nsplits |
the number of splits to make in calculating split-half reliability |
quest.reliability
will only handle numeric scores. Character
scores can be transformed into numeric using as.numeric(as.factor(...
but take care that the transformation preserves the order of the scores.
Any rows with NA values in the score matrix will be removed before the
calculations are made.
A list containing five components:
cronbach.alpha |
Cronbach's alpha for the entire scale. |
split.half |
An approximate split-half reliability is
calculated by dividing the items into halves on a random basis and
calculating the correlation between the summed scores for the halves.
This is repeated nsplits times. |
item.whole |
The correlation of each item with the sum of the remaining items is calculated. |
item.alpha |
Cronbach's alpha is calculated for each item on all other items, i.e. the scale alpha if that item were dropped. |
item.var |
The variance of scores for each item is calculated. Items with relatively low variance are likely to be poor discriminators. |
Jim Lemon
# fake a data set with one bad item fakedata<-matrix(c(rep(c(2,2:9,9),9)+sample(c(-1,0,1),90,TRUE), rep(5,10)+sample(c(-1,0,1),10,TRUE)),nrow=10) quest.reliability(fakedata)