Two-phase sampling fixes (#42)

* fitting mechanisms are now only fit to appropriate subsets of the data

* updating two-phase sampling tests

* addressing minor bugs, passing checks

* loosening nuisance mse test conditions

* fixing TML estimator weighting scheme

* eif reordering in one step occurs only if required

* TML and one step estimators re-roder EIF

* fixing two-phase indexing issue in tml estimator

* tilting two-phase weights

* fixing issue with data.table creation during weight tilting

* two-phase sampling weight tilting occurs iteratively

* removing two-phase weight truncation, hajek stabilization instead

* renaming variables used to compute PM

* two-phase sample weights are now stabilized and truncated

* reverting to iterative truncation strategy

* increasing truncation in tilting procedure

* weighted tilting for two-phase weights

* contrast information for pm estimators diagnostics

* fixed compounding weights in two-phase prob tilting

* only two-phase weights are used for TMLE tilting in two-phase sampling schemes

* fix final out reg tilting in two-phase sampling

* updating seeds to make sure stochastic tests pass

* revert to previous full-data EIF tilting approach

* always nudge the b nuisance parameter estimate

* don't use tilt_stop_cript

* adding weights to b_score

* removing tilting information from output

* adding two-phase weights to clever covariates

* drop two-phase weights from plugin in TML

* v tilting mechanism includes two_phase_weights now

* avoid over tilting two-phase sampling weights

* touch ups to TMLE before review

* tweaks to weighting scheme

* two-phase weights to glm weights for TML tilting

---------

Co-authored-by: Nima Hejazi <nh@nimahejazi.org>

DESCRIPTION

@@ -66,4 +66,4 @@ BugReports: https://github.com/nhejazi/medoutcon/issues
 Encoding: UTF-8
 LazyData: true
 VignetteBuilder: knitr
-RoxygenNote: 7.2.1
+RoxygenNote: 7.2.2