binest

Estimation of group-level means and standard deviations from binned (coarsened) count data. All three functions fit the same heteroskedastic ordered probit (HETOP) model, in which each group’s values are normally distributed around a mean and SD of its own. They differ only in how they fit it, and they share a common output structure:

This package was previously called HETOP and was maintained by J. R. Lockwood; it is renamed and extended to reflect the broader functionality now included.

Installation

# install.packages("remotes")
remotes::install_github("paulvonhippel/binest")

(Or install.packages("binest") for the released version on CRAN.)

A quick example

library(binest)
data(tx_g6_math_2018)

ngk <- with(tx_g6_math_2018,
            cbind(unsatisfactory, approaches, meets, masters))
cuts <- c(1536, 1653, 1772)

## `scope` is required: Texas reports counts for every tested student,
## so each district's students are its whole population and the only
## uncertainty is the binning. Use scope = "sample" when the units are
## a sample from a larger population, which adds sampling error to the
## reported SEs.
fit <- fast_hetop(ngk, cutpoints_known = TRUE, cutpoints = cuts,
                  scope = "population")
cor(fit$est_raw$mean, tx_g6_math_2018$reported_mean)

See vignette("binest") for a full comparison of the three functions on the Texas STAAR Grade 6 mathematics data.

References

License

GPL (>= 2).