Description Usage Arguments Details Value References
An interactive tool for estimating principal stratification-based causal effects complete with sensitivity analysis. User can define distributions for factual and counterfactual trial data and compute causal effect and 95% confidence intervals. Data set can be loaded at function call or loaded later by interacting with the tool. Tool provides feedback regarding distributional assumptions that would give rise to invalid causal effect estimates. By varying the sensitivity parameter (S(1)|Y(0)=1 distribution), the user can examine the robustness of the causal effect estimate.
1 |
data |
Optional dataset. File format requirements are:
|
meanS1 |
Optional estimated mean for biomarker value for treatment group. Ignored if a dataset is assigned to |
sdS1 |
Optional estimated standard deviation for biomarker value for treatment group. Ignored if a dataset is assigned to |
Currently supported distributions for the factual S(1) biomarker data in the treatment group:
Normal: Parameters are estimated from the data's mean and standard deviation.
Gamma: Parameters are estimated from the data's mean and standard deviation. If the data would yield an invalid estimate for either the gamma distributions scale or shape parameters, an error message is given.
Non-parametric: Kernel density estimating using density
.
For the sensitivity parameter S(1)|Y(0)=1, slider bars are provided to allow the user to vary the mean and the standard deviation of the sensitivity parameter's distribution. The slider bars are in reference to the mean and standard deviation for S(1). Supported distributions include:
Normal: Parameters are estimated from the parameter's mean and standard deviation, which are scaled from those of S(1) using the slider bars.
Gamma: Parameters are estimated from the parameter's mean and standard deviation, which are scaled from those of S(1) using the slider bars. If the data would yield an invalid estimate for either the gamma distributions scale or shape parameters, an error message is given.
Non-parametric: This option uses the non-parametric density estimation for S(1) offset by the estimated mean for S(1)|Y(0) from the slider bars. S(1) must be set to “non-parametric” for this option to be valid.
The tool provides a histogram of the S(1) data and a density sketch for S(1) using current assumptions.
The tools also displays the density assumption for the sensitivity parameter S(1)|Y(0) and its reference distribution, S(1), in the plot marked “Distribution of S(1)|Y(0)=1”.
Causal effect is estimated and displayed in the plot marked “Causal Effect (Given S(1))”.
Confidence intervals (95%) optionally be calculated using boot
and displayed on the
causal effect plot.
The tool also provides the capability to zoom in on a section of the causal effect plot. When it zooms, only the causal effect is displayed. Any confidence intervals must be recalculated.
The widgets that comprise the tool are from rpanel.
Known bug:
The “Load Data File” function does not always exit cleanly when attempting to load a file with an unexpected format.
No value is returned by the function.
Bowman, Adrian et al., “rpanel: Simple Interactive Controls for R Functions Using the tcltk Package”, Journal of Statistical Software, Jan 2007, Vol 17, Issue 9. See also http://www.stats.gla.ac.uk/~adrian/rpanel/.
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.