binest: Estimation of Group Means and SDs from Binned Count Data
Estimates group-level means and standard deviations from
binned (coarsened) data, where only the number of cases in each of
several ordered bins is observed and the underlying values are not.
The main function, fast_hetop(), fits the heteroskedastic ordered
probit model of Reardon, Shear, Castellano and Ho (2017)
<doi:10.3102/1076998616666279> one group at a time, solving
closed-form truncated-normal score equations instead of optimizing
jointly over all groups. Runtime is therefore linear in the number
of groups, so thousands of schools, districts or subgroups can be
fitted in under a second. Cut scores may be supplied when an
agency publishes them, or estimated from the data when it does not.
Output includes standard errors and confidence intervals, an
optional empirical Bayes shrinkage estimator, and a per-group
goodness-of-fit test of the within-group normality assumption.
Also included, and deprecated, are mle_hetop() and fh_hetop(),
which fit the same model by joint maximum likelihood and by Markov
chain Monte Carlo (Lockwood, Castellano and Shear 2018
<doi:10.3102/1076998618795124>); they are forked from the 'HETOP'
package by J. R. Lockwood and retained only for comparison.
Documentation:
Downloads:
Linking:
Please use the canonical form
https://CRAN.R-project.org/package=binest
to link to this page.