Health Outcomes Software Downloads

Go Back

Unsupervised & Supervised Propensity Scoring: USPS in R

Download Windows "Zip" Package, Version_1.01-1

Download Linux "Tar.Gz" Package, Version_1.01-1

September 2006

 

UPGRADE 2006: Bob Obenchain's R Package for both Unsupervised and Supervised Propensity Scoring has a new function for computing and displaying the "Artificial LTD Distribution" due to random patient clusterings. When comparisons are made within highly relevant patient clusters in covariate X-space, the resulting Local Treatment Difference (LTD) distributions are usually much different.  Supervised methods are also provided for spline or loess "smoothing" across traditional PS bins defined using logistic regression models.  Extensive *.pdf and *.html documentation files are included within the archives.

 

"Local Control" and "Artificial LTD Distribution" Scripts for JMPÒ

Download "Zip" Archive

July 2006
 

Bob Obenchain's JMP scripts for Nearest Neighbors LTD methodology and Artificial LTD Distributions (from random clustering) in Unsupervised Propensity Scoring.  This implementation is highly interactive and visually-oriented and features the most important methods from my USPS package for R (but neither IV nor Supervised PS methods.)  Extensive PDF documentation, white papers, references and example datasets are included within the distribution archive.

 

Download "PSbinbst" SetupFactory Executable,

Version 2002.03

[462K]
 

This is the March 2002 revision of Bob Obenchain's console application for Propensity Score Bin BootStrapping. of a single outcome. Starting with data from two independent samples of patients, the user of PSbinbst must identify the outcome for each patient by [1] the treatment he/she received (coded 0=new or 1=standard) and [2] that patient's "bin number" from 1 to N ...where N is at least 2 and at most 8.  (These sorts of bins can be created using functions from my Propensity Scoring in R archive.)

Incremental Cost-Effectiveness (ICE) Statistical Inference

 

Download "ICEplane" SetupFactory Executable,

Version 2005.11

[1741K]
UPDATE: This is the November 2005 revision of Bob Obenchain's "ICEplane" software for making Incremental Cost-Effectiveness (ICE) comparisons between two independent, unbiased samples of patients.   Five benchmark datasets from ICE literature are included to support software validation efforts.  ICEplane is a Microsoft Windows application that uses Graphics Server® to manage graphical displays.  ICEplane creates Confidence regions (based on bootstrap resampling or Fieller's theorem), Tolerance regions, Acceptability Curves and (linear) Net Benefit analyses.  The user of ICEplane must specify a shadow price of health for converting the effectiveness measure into a cost or vice versa.  Archive also contains C-language source code and an executable for the 1998 version of ICERconf.EXE.

Download "ICEpairs" SetupFactory Executable,

Version 2002.09

[1101K]
This is the September 2002 revision of Bob Obenchain's "ICEpairs" software for making within-patient cost-effectiveness comparisons from cross-over study data.  This Microsoft Windows application uses Graphics Server® to manage graphical displays.  ICEpairs creates Confidence regions, Tolerance regions and Acceptability Curves for Incremental Cost Effectiveness (ICE) statistics.  Download includes results from an illustrative "Hands-On, Lips-On ICE" workshop.

 

Download "ICEpsbbs" SetupFactory Executable,

Version 2002.05

[947K]
This is the May 2002 revision of Bob Obenchain's "ICEpsbbs" software for ICE Propensity Score Bin BootStrapping.  Starting with data from two independent (possibly biased) samples of patients, the user of ICEpsbbs must identify the cost-effectiveness outcome pair for each patient by [1] the treatment he/she received (coded 0=new or 1=standard) and [2] that patient's "bin number" from 1 to N ...where N is at least 2 and at most 8.  (These sorts of bins can be created using functions from my Propensity Scoring in R archive.)  This Microsoft Windows application uses Graphics Server® to manage graphical displays.  ICEpsbbs creates Confidence regions, Tolerance regions, Acceptability Curves and Net Benefit analyses of Incremental Cost Effectiveness (ICE) statistics. 

 

Download "EL_GAUSS.ZIP" Archive,
Version 2005.12

[128K]
Download the most recent versions of Bob Obenchain's GAUSS language functions (*.g) and examples (*.e) using the optional Constrained Optimization (CO) library to identify convex Empirical Likelihood confidence regions in ICE statistical inference.  Archive contains PDF files of examples.

 

Download "ICEuncrt" Self-Extracting Archive,
Version 2K06

[103K]
Download the June 2000 version of Bob Obenchain's 32-bit Microsoft Windows "Console Application" that simulates correlated, bivariate normal cost and effectiveness outcomes to illustrate the one-to-one relationship between ICE radius and overall precision.

Download "CE_Mult" Self-Extracting Archive, Version 9805

[220K]
Download Bob Obenchain's self-extracting archive for his May 1998 "CE_Mult" 32-bit Microsoft Windows "Console Application" that simulates multiplicity bias.  This is the bias introduced into ICE confidence intervals (and regions) when several possible cost or effectiveness measures are interchangeable but only the most-favorable result (over choice of numerator and denominator) is reported

Simulation of Informatively Censored Outcomes

Download SurvCest ver.2K03 Archive

[163K]
Download the March 2000 version of Bob Obenchain's 32-bit Microsoft Windows "Console Application" for generating informatively censored outcomes (disease duration and cost) via a 3-state Markov chain model.  Patients can only "discontinue" when they are in the most severely ill state ...i.e. when they are in the state with highest expected duration until cure and where costs are being accumulated at the highest expected rate.  Kaplan-Meier estimates of cost survival curves (time = money) can be badly biased downwards in these cases!

 

Go Back