plot_profile: Visualization for Model Selection/Tuning

In jedazard/PRIMsrc: PRIM Survival Regression Classification

Visualization for Model Selection/Tuning

Description

Function for plotting the cross-validated model selection/tuning profiles of a sbh object. It uses the user's choice of cross-validation criterion statistics among the Log Hazard Ratio (LHR), Log-Rank Test (LRT) or Concordance Error Rate (CER). The function plots (as it applies) both profiles of cross-validation criterion as a function of variables screening size (cardinal subset of top-screened variables in the PRSP variable screening procedure), and peeling length (number of peeling steps of the peeling sequence in the inner loop of the PRSP or PRGSP algorithm).

Usage

  plot_profile(object,
               main = "Profile Plots",
               xlim = NULL,
               ylim = NULL,
               add.sd = TRUE, 
               add.profiles = TRUE,
               add.caption = TRUE, 
               text.caption = c("Mean","Std. Error"),
               pch = 20, 
               col = 1, 
               lty = 1, 
               lwd = 0.5, 
               cex = 0.5,
               device = NULL, 
               file = "Profile Plots", 
               path = getwd(), 
               horizontal = FALSE, 
               width = 8.5, 
               height = 5.0, ...)

Arguments

object	Object of class sbh as generated by the main function sbh.
main	Character vector. Main Title. Defaults to "Profile Plots".
xlim	Numeric vector of length 2. The x limits [x1, x2] of the plot. Defaults to NULL.
ylim	Numeric vector of length 2. The y limits [y1, y2] of the plot. Defaults to NULL.
add.sd	Logical scalar. Shall the standard error bars be plotted? Defaults to TRUE.
add.profiles	Logical scalar. Shall the individual profiles (for all replicates) be plotted? Defaults to TRUE.
add.caption	Logical scalar. Should the caption be plotted? Defaults to TRUE.
text.caption	Character vector of caption content. Defaults to {"Mean","Std. Error"}.
pch	Integer scalar of symbol number for all the profiles. Defaults to 20.
col	Integer scalar of line color of the mean profile. Defaults to 1.
lty	Integer scalar of line type of the mean profile. Defaults to 1.
lwd	Numeric scalar of line width of the mean profile. Defaults to 0.5.
cex	Numeric scalar of symbol expansion for all the profiles. Defaults to 0.5.
device	Graphic display device in {NULL, "PS", "PDF"}. Defaults to NULL (standard output screen). Currently implemented graphic display devices are "PS" (Postscript) or "PDF" (Portable Document Format).
file	File name for output graphic. Defaults to "Profile Plot".
path	Absolute path (without final (back)slash separator). Defaults to working directory path.
horizontal	Logical scalar. Orientation of the printed image. Defaults to FALSE, that is potrait orientation.
width	Numeric scalar. Width of the graphics region in inches. Defaults to 8.5.
height	Numeric scalar. Height of the graphics region in inches. Defaults to 5.0.
...	Generic arguments passed to other plotting functions.

Details

Model tuning is done by applying the cross-validation criterion defined by the user's choice of specific statistic. The goal is to find the optimal value of model parameters by maximization of LHR or LRT, or minimization of CER. The parameters to optimize are (i) the cardinal of top-ranked variables subsets (if the "prsp" variable screening is chosen), and (ii) the number of peeling steps of the peeling sequence (inner loop of our PRSP algorithm) in any case of variable screening method.

Currently, this is done internally for visualization purposes, but it will ultimately offer the option to be done interactively with the end-user as well for parameter choosing/model selection.

Value

Invisible. None. Displays the plot(s) on the specified device.

Acknowledgments

This work made use of the High Performance Computing Resource in the Core Facility for Advanced Research Computing at Case Western Reserve University. This project was partially funded by the National Institutes of Health NIH - National Cancer Institute (R01-CA160593) to J-E. Dazard and J.S. Rao.

Note

End-user plotting function.

Author(s)

• "Jean-Eudes Dazard, Ph.D." [e-mail]

• "Michael Choe, M.D." [e-mail]

• "Michael LeBlanc, Ph.D." [e-mail]

• "Alberto Santana, MBA." [e-mail]

• "J. Sunil Rao, Ph.D." [e-mail]

Maintainer: "Jean-Eudes Dazard, Ph.D." [e-mail]

References

• Dazard J-E. and Rao J.S. (2021a). "Variable Selection Strategies for High-Dimensional Recursive Peeling-Based Survival Bump Hunting Models." (in prep).

• Dazard J-E. and Rao J.S. (2021b). "Group Bump Hunting by Recursive Peeling-Based Methods: Application to Survival/Risk Predictive Models." (in prep).

• Dazard J-E., Choe M., Pawitan Y., and Rao J.S. (2021c). "Identification and Characterization of Informative Prognostic Subgroups by Survival Bump Hunting." (in prep).

• Rao J.S., Huilin Y., and Dazard J-E. (2020). "Disparity Subtyping: Bringing Precision Medicine Closer to Disparity Science." Cancer Epidemiology Biomarkers & Prevention, 29(6 Suppl):C018.

• Yi C. and Huang J. (2017). "Semismooth Newton Coordinate Descent Algorithm for Elastic-Net Penalized Huber Loss Regression and Quantile Regression." J. Comp Graph. Statistics, 26(3):547-557.

• Dazard J-E., Choe M., LeBlanc M., and Rao J.S. (2016). "Cross-validation and Peeling Strategies for Survival Bump Hunting using Recursive Peeling Methods." Statistical Analysis and Data Mining, 9(1):12-42.

• Dazard J-E., Choe M., LeBlanc M., and Rao J.S. (2015). "R package PRIMsrc: Bump Hunting by Patient Rule Induction Method for Survival, Regression and Classification." In JSM Proceedings, Statistical Programmers and Analysts Section. Seattle, WA, USA. American Statistical Association IMS - JSM, p. 650-664.

• Dazard J-E., Choe M., LeBlanc M., and Rao J.S. (2014). "Cross-Validation of Survival Bump Hunting by Recursive Peeling Methods." In JSM Proceedings, Survival Methods for Risk Estimation/Prediction Section. Boston, MA, USA. American Statistical Association IMS - JSM, p. 3366-3380.

jedazard/PRIMsrc documentation built on July 16, 2022, 10:56 p.m.