This PR adds a few new features, additional unit tests, and bug fixes.
Added 'stddev" as an additional performance metric under calc_absolute(), per request from Luke Miratrix (#21).
Added a winz option to calc_absolute(), calc_relative(), calc_coverage(), and calc_relative_var() to do winsorization of the sampling distribution prior to calculating performance criteria. This is one simple (but kind of kludgy) way to handle outlying point estimates (or outlying standard error estimates in calc_relative_var()). It's disabled by default, with winz = Inf. The typical winsorization constant would be 1.5.
Reworked calc_relative() to call calc_absolute() and use the result to calculate the relative performance measures. This reduces the amount of repeated code.
Added a enclos argument to the eval() calls inside all the calc_*() functions. This allows the functions to find values of true_param or other quantities that are not on the specified data.
Additional unit tests for the winz argument and to validate calc_relative_var() against the results of calc_absolute() and calc_relative().
I'd like to use these updates in the selection model sims. Could you please review and pull in to main branch when you have time?
This PR adds a few new features, additional unit tests, and bug fixes.
'stddev"
as an additional performance metric undercalc_absolute()
, per request from Luke Miratrix (#21).winz
option tocalc_absolute()
,calc_relative()
,calc_coverage()
, andcalc_relative_var()
to do winsorization of the sampling distribution prior to calculating performance criteria. This is one simple (but kind of kludgy) way to handle outlying point estimates (or outlying standard error estimates incalc_relative_var()
). It's disabled by default, withwinz = Inf
. The typical winsorization constant would be 1.5.calc_relative()
to callcalc_absolute()
and use the result to calculate the relative performance measures. This reduces the amount of repeated code.enclos
argument to theeval()
calls inside all thecalc_*()
functions. This allows the functions to find values oftrue_param
or other quantities that are not on the specifieddata
.winz
argument and to validatecalc_relative_var()
against the results ofcalc_absolute()
andcalc_relative()
.I'd like to use these updates in the selection model sims. Could you please review and pull in to main branch when you have time?