Plots the root mean squared error (RMSE) or bias of estimated parameters as a function of their true (simulated) values, aggregated into bins. This reveals how estimation accuracy varies across the parameter range (e.g., whether extreme item difficulties are estimated less precisely).
Accepts output from [optim_sample()], [get_sspLNIRT()], or [comp_rmse()]. For [optim_sample()] output, the error data at the minimum \(N\) are used. If the data are already binned (i.e., `keep.err.dat = FALSE` in [comp_rmse()]), bins are plotted as-is; otherwise, raw errors are binned on the fly using `n.bins`. For item parameters, \(\sigma^2\) is excluded.
This function is preserved for backward compatibility. New code should prefer `plot(object, type = "estimation", ...)`.
Arguments
- object
An object of class `"sspLNIRT"`, as returned by [optim_sample()], [get_sspLNIRT()], or [comp_rmse()].
- pars
Character. `"item"` or `"person"`. Which parameter set to plot.
- y.val
Character. `"rmse"` or `"bias"`. Metric for the y-axis.
- n.bins
Integer. Number of quantile bins for aggregation. Only used when the error data are in full (unbinned) format. Default is 30.
See also
[plot.sspLNIRT()] for the recommended interface; [plot_power_curve()] for visualizing the optimization trace; [theme_sspLNIRT()].
Examples
if (FALSE) { # \dontrun{
plot_estimation(result, pars = "item", y.val = "rmse")
# equivalent and preferred:
plot(result, type = "estimation", pars = "item", y.val = "rmse")
} # }