NEWS.md

In forsearch: Diagnostic Analysis Using Forward Search Procedure for Various Models

1/7/2025 forsearch version 6.4.0

• Major changes: None

• Moderate changes:

• Added vignette on Study Credibility
• Added arguments to forsearch_nls to liberalize controls

• Added function logist3 to calculate 3-parameter logistic

• Minor changes and bug fixes:

• Removed phases in forsearch_nls in favor of sections
• Returned to use of stats::predict for all forsearch functions
• Updated all examples

11/15/2024 forsearch version 6.3.0

• Major changes: None

• Moderate changes:

• Added candprep support function
• Optimized step 1 of every forsearch_foo function
• Identified observation set within step 1 as yielding lowest median squared error WITHIN each factor subset
• Changed to 2-phase procedure in step 2 to ensure all parameters can be estimated at every stage in step 2
• Added arguments to forsearch_lme to liberalize controls

• Changed from predict to predict.lme, predict.glm and predict.coxph in xStep1 and xStep2

• Minor changes and bug fixes:

• Modified showme to limit output for forsearch_lme
• Added graphs for forsearch_lme
• Corrected use of deviances in forsearch_glm
• Removed forsearch_nls for corrections and improvements

7/13/2024 forsearch version 6.2.0

• Major changes: None

• Moderate changes:

• Revised aStep1 to improve step 1 for _lm

• Added bStep1 and cStep1 to improve step 1 for _lme and _cph

• Minor changes and bug fixes:

• Added option to disturb event times slightly to avoid duplicates in forsearch_cph observations
• Changed message in plotdiag.ANOX2 if proportionality not tested in proportional hazard data
• Minor text changes to vignettes
• Corrections to examples to correspond to function changes

6/22/2024 forsearch version 6.1.0

• Major changes: None

• Moderate changes:

• Enabled models with constructed variables (those that appear in the formula but not in the database)

• Minor changes and bug fixes: None

3/30/2024 forsearch version 6.0.0

• Major changes:

• Added functions forsearch_nls, eStep1, and eStep2 for nonlinear statistics

• Moderate changes: None

• Minor changes and bug fixes:

• Added code to function showme for forsearch_nls objects and to function plotdiag.allgraphs for forsearch_nls graphics
• Modified vignette regarding where we get observations to cover nonlinear models

2/16/2024 forsearch version 5.1.0

• Major changes: None

• Moderate changes:

• Completely revised flow of forsearch_glm to better accommodate models containing independent variables as factors

• Added vignette regarding exploration of the search history

• Minor changes and bug fixes:

• Fixed proportionality test bug (na's) in plotdiag.ANOX2 for Cox regression
• Changed use of n.obs.per.level in picksome to be additive
• Remove unused argument of bStep2
• Added blind.label to arguments of plotdiag.allgraphs to allow user to request labeling of graph to reflect blinded database

1/13/2024 forsearch version 5.0.0

• Major changes:

• Completely revised flow of forsearch functions _lm, _lme and _cph to better accommodate models containing independent variables as factors

• Moderate changes:

• Added vignette describing strategy of the forsearch procedure

• Added ability to skip test of proportionality in forsearch_cph

• Minor changes and bug fixes:

• Modified diagnostic printouts for forsearch_cph
• Removed option in variablelist and in picksome for identifying function
• Integrated diagnostic printouts of aStep1, aStep2 and bStep2 with those of calling functions
• Added requirement that forsearch_cph have at least 1 factor variable
• Fixed bug preventing use of multiple observations in Step 1.
• Cleaned up examples

11/6/2023 forsearch version 4.2.0

• Major changes: None

• Moderate changes:

• Added plotdiag.ANOX2 and applied it to proportionality test output of forsearch_cph, and test of null hypothesis in forsearch_lm and forsearch_lme
• Increased scope of models that can by handled by forsearch_lme by revising forsearch_lme and bStep2. Removed need for bStep1

• Added vector of Step 1 observation numbers to all forsearch_xxx functions

• Minor changes and bug fixes:

• Added proportionality test (cox.zph) to forsearch_cph
• Added test of null hypothesis (anova) to forsearch_lm and forsearch_lme
• Added proportionality test to showme for forsearch_cph
• Added null hypothesis test to showme for forsearch_lm and forsearch_lme
• Correct failure to define response.colnum in forsearch_lme that prevented skip of Step1
• Changed argument "addline" in plotdiag.Cook to single value default

9/12/2023 forsearch version 4.1.0

• Major changes: None

• Moderate changes: None

• Minor changes and bug fixes:

• Replaced argument diagnose=T/F with begin.diagnose numeric indicator
• Removed argument diagnose and diagnostic code in several small functions
• Added examples in all forsearch_xxx functions
• Cleaned up documentation wording

8/8/2023 forsearch version 4.0.0

• Major changes:
• Added function forsearch_cph for Cox proportional hazards
• Added function cStep2 for step 2 of Cox proportional hazard function

• Added functions plotdiag.Wald, plotdiag.lrt, and plotdiag.loglik

• Moderate changes:

• Combined showme, showmegl, and showmelme into showme function

• Added forsearch_cph to showme function

• Minor changes and bug fixes:

• Correct display of ANOVA structure in forsearch_glm for family=Gamma
• Correct indexing in aStep1
• Correct function ID in plotdiag.leverage
• Set default for addline in plotdiag.s2
• Added message in variablelist regarding single observations in some factor levels
• Correct coding error in picksome function

2/17/2023 forsearch version 3.2.0

• Major changes: None

• Moderate changes:

• Added guidance to prevent linearly dependent X'X matrices in forsearch_glm

• Added capability of forsearch_glm to handle ANOVA and ANCOVA models

• Minor changes and bug fixes:

• Correct coding error in picksome function

1/4/2023 forsearch version 3.0.1

• Major changes: None

• Moderate changes: None

• Minor changes and bug fixes:

• Fixed response column identifier in forsearch_lme, aStep1, aStep2, bStep1 and bStep2
• Fixed need for handling NaN in certain data for forsearch_glm
• Clarified help for forsearch_glm for two response column numbers

12/14/2022 forsearch version 3.0.0

• Major changes:
• Rewrote forsearch_lme to clarify code for generation of Steps 1 and 2
• Extracted Step 1 and Step 2 increases in sample size to separate functions aStep1 and aStep2 from forsearch_lm and forsearch_lme
• Added functions aStep1 and aStep2 for linear model functions

• Added functions bStep1 and bStep2 for linear mixed effects models

• Moderate changes:

• Added plotdiag.all function to create drafts of all applicable graphs
• Added functions variablelist() and picksome() to subdivide observations into their factor levels
• Added capability of forsearch_lm, forsearch_glm and forsearch_lme to process factor variables and interactions with regression variables
• Added capability of forsearch_lm to calculate leverage of observations over subsets

• Added examples for forsearch_lm, forsearch_glm, and forsearch_lme

• Minor changes and bug fixes:

• Changed README to add plotdiag.all, variablelist, and picksome functions
• Changed some plotting functions to allow no fit of a line to the data
• Added option for all forsearch_xxx functions to blind initial display of data analysis structure, anticipating future developments

4/21/2022 forsearch version 2.3.0

• Major changes: None

• Moderate changes:

• Correct subgroup selection procedure in forsearch_lme
• Calculate root mean square of random coefficients instead of the coefficients themselves in forsearch_lme
• Correct plotdiag.params.random to correspond with changed calculation of RMS

• Add plotdiag.fit3 to plot AIC, BIC, and log likelihood for forsearch_lme objects

• Minor changes and bug fixes:

• Print summary of assumed analysis in forsearch_lm and in forsearch_lme to permit formulation check
• Modify README and documentation of functions

3/24/2022 forsearch version 2.2.0

• Moderate changes:
• Clean up README and help files
• Modify forsearch_glm to handle observations of 0 or N (binomial)

3/17/2022 forsearch version 2.1.0

• Moderate changes:
• Correct the omission of README file

3/16/2022 forsearch version 2.0.0

• Major changes:
• Added forsearch_glm() function
• Added plotdiag.deviance.residuals() function
• Added plotdiag.phihatx() function
• Added plotdiag.AICX() function

• Added plotdiag.deviances() function

• Moderate changes:

• Added the collection of t statistics in forsearch_lm() and forsearch_lme() for subsequent plotting
• Added plotdiag.tstats() function to plot t statistics
• Added showmegl() function

• Changed some existing plotdiag functions to accommodate forsearch_glm()

• Minor changes and bug fixes:

• Added t statistics to list of summary prints in showme() and showmelme()
• Modified search.history() to account for case when all current observations replaced by new ones
• Enabled user to enter caption on all graphs
• Corrected 2nd example in README file
• Updated README file
• Removed extra squaring of response in plotdiag.residuals() function

1/27/2022 forsearch version 1.0.0

First version of this package

forsearch documentation built on April 4, 2025, 5:52 a.m.