Test the adequacy of simple choice, logistic, or Thurstonian scaling.

Given a matrix of choices and a vector of scale values, how well do the scale values capture the choices? That is, what is size of the squared residuals given the model versus the size of the squared choice values?

scaling.fits(model, data, test = "logit", digits = 2, rowwise = TRUE)

Arguments

model: A vector of scale values
data: A matrix or dataframe of choice frequencies
test: "choice", "logistic", "normal"
digits: Precision of answer
rowwise: Are the choices ordered by column over row (TRUE) or row over column False)

Details

How well does a model fit the data is the classic problem of all of statistics. One fit statistic for scaling is the just the size of the residual matrix compared to the original estimates.

Value

GF: Goodness of fit of the model
original: Sum of squares for original data
resid: Sum of squares for residuals given the data and the model
residual: Residual matrix

References

Revelle, W. (in preparation) Introduction to psychometric theory with applications in R, Springer. https://personality-project.org/r/book/

Author

William Revelle

Note

Mainly for demonstration purposes for a course on psychometrics