To use the program, download all of these files: sim.R , gridExtensions.R , graphInput.R , manageData.R , toolBar.R . Then start R, and use the menu item "File/Change dir..." to navigate to the directory containing the downloaded files. Then source "sim.R" using the menu item "File/Source R Code..." or the source command.

(The easiest way to work in R under Windows is to create a SimTool project directory with the downloaded files. Then add a shortcut to R with " --sdi" at the end of the Target line and with the correct directory in the StartIn line. Then R won't steal the entire desktop, and also you won't need to use "File/Change dir...". In future sessions, you also don't need to re-source any files.)

To work with a particular simulation scenario, enter some version of the command sim("MyScenario"). If the scenario is new, you will be given a chance to customize certain basic aspects of the scenario. All other aspects may be changed at any time.

After the initial customization, enter "q" to quit, and you will see the main interaction screen in a graphics window. All further interactions are mouse driven and in this screen. Choose a main task with the tool bar at the top of the screen. The choices are:

• Subjects: modify number of subjects and demographics
• Study: modify study length and model parameters
• Simulate: simulate a single study and see EDA and Cox PH statistics
• Sim x10: simulate 10 studies automatically (there is a study option which sets whether or not the program pauses and waits for a click after each simulation)
• EDA: Repeats the EDA for the most recent study
• Samp. Dist: Shows the sampling distribution of the drug parameter in the Cox PH model
• Save: save the latest simulation data to a ".csv" file for analysis in another program.
• Quit

The simulation EDA and statistics are as follows. In the upper left is a cross-tabulation by gender and treatment assignment. In the upper left corner of this table is the simulation number, which resets to zero when any settings affecting the nature of the simulation are changed. Also on the top portion of the EDA are histograms of age and disease severity. The lower left portion of the EDA shows the Kaplan Meyer survival curves for the two treatments (ignoring age, gender and severity). The lower right portions shows results from a Cox Proportional Hazards model with treatment, age, gender and severity as the explanatory variables. The coefficient column shows the estimated change in log odds of survival for a one unit change in the explanatory variable. The OR column shows this value exponentiated which is the odds ratio for a one unit change. The p-value are also shown. Below the table is the overall likelihood ratio test result. If the number of simulations is at least 10, the observed power for detected a treatment effect is also shown.

The Sampling Distribution of the Cox PH "treatment" parameter shows how that value would change over repeat runs of the experiment. By chance your one run of the experiment could turn out to be any one of those values.

The Save button saves the demographics, treatment assignment, plus "relapse" and status into a "commas separated values" file, which is compatible with most statistical packages. The "relapse" variable is either the time of relapse or the time of censoring. If "status" is 0, the relapse time is a censoring time; if it is 1, the relapse time is a true relapse time.

Here is the initial screen, showing the tool bar and current settings:

Here is the subject screen, where subject setting may be altered, and (ideal) subject characteristics are graphed:

Here is the study screen, where study setting may be altered, and (ideal) recurrence rates are graphed:

Here is the EDA/simulation screen, where simulation results are presented:

Here is the sampling distribution screen, which shows a histogram of the treatment coefficient estimates over repeated simulations:

All links active 4/13/2005. Please report missing links to hseltman at stat.cmu.edu.

To my Home Page