plot-MessinaSurvResult-missing-method: Plot the results of a Messina analysis on a survival problem.

Description Usage Arguments Details Author(s) See Also Examples

Description

Plots diagnostic and performance information for fits in a MessinaSurvResult object, as returned by messinaSurv.

Usage

1
2
## S4 method for signature 'MessinaSurvResult,missing'
plot(x, y, ...)

Arguments

x

the result of a Messina survival analysis, as returned by messinaSurv.

...

additional options to control the plot:

indices

a vector of indices of features to plot. If sort_features == FALSE, the indices are into the unsorted features, as originally supplied in x supplied to messinaSurv. If sort_features == TRUE, features are first sorted in order of decreasing margin, and then the indices in this parameter are plotted. For example, if indices == 2 and sort_features == FALSE, the second feature in x will be plotted. However, if sort_features == TRUE, the feature with the second best classifier margin will be plotted.

sort_features

a boolean indicating whether to sort features by decreasing margin size before selecting from indices. This affects the interpretation of the parameter 'indices'; for more details see the description of that parameter.

bootstrap_type

a string giving the type of bootstrap error band to produce on the survival prediction plots. Can take three values: "none", "stdev", and "ci". "none", the default, plots no error bands. "stdev" performs multiple rounds of Kaplan-Meier curve estimation on bootstrap samples, and plots prediction bands corresponding to +/- 1 bootstrap standard deviation from the mean. "ci" performs bootstrapping as per "stdev", and plots prediction bands corresponding to the bootstrap_ci intervals.

bootstrap_ci

a value in (0.5, 1) giving the confidence interval for bootstrap_type == "ci". Ignored otherwise. Default 0.9 for 90% confidence intervals.

nboot

the number of bootstrap iterations to perform for calculations. Set to a reasonable default taking into account bootstrap_type and bootstrap_ci, so ordinarily does not need to be specified by the user.

parallel

a logical indicating whether multiprocessing using doMC should be used for the bootstrap calculations. If NULL, multiprocessing will be used if doMC is loaded and more than one parallel worker is registered.

y

the y coordinates of points in the plot, optional if x is an appropriate structure.

Details

For each feature index given by indices, produces four plots:

"Objective function"

A plot of the value of the objective function over all possible thresholds. Each sample is represented by a point on the objective function trace. The selected threshold, if any, is shown by a solid vertical line, and the margins by dotted vertical lines on either side of it. The minimum values of the objective function specified by the user are shown as horizontal dotted lines. This plot is useful for assessing fit stability, particularly for the "coxcoef" and "reltau" objective functions, which can be unstable at low or high threshold values. See messinaSurv for details.

"Separation performance at threshold"

This Kaplan-Meier plot shows two traces, showing the outcomes of the two subgroups in the cohort defined by whether the plotted feature is above or below the threshold. Optionally (if bootstrap_type != "none"), the KM traces will be surrounded by shaded regions that represent either +/- 1 SD (bootstrap_type == "stdev") or a bootstrap_ci confidence interval (bootstrap_type == "ci").

"Separation performance at lower margin"

This plot is identical to the above, except that the performance when the lower margin is used to separate the sample groups is shown.

"Separation performance at lower margin"

This plot is identical to the above, except that the performance when the upper margin is used to separate the sample groups is shown. These last two plots give an indication of the robustness of the MessinaSurv fit at its extremes.

The Kaplan-Meier plots may optionally display bootstrap bands, if bootstrap_type != "none". Note that the calculation of bootstrap bands is computationally-intensive, and this function will by default use multiprocessing to speed calculations if doMC is loaded and more than one core registered for use. For examples of the plots and their interpretation, see the vignette.

Author(s)

Mark Pinese [email protected]

See Also

MessinaSurvResult-class

messinaSurv

Examples

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
## Load a subset of the TCGA renal clear cell carcinoma data
## as an example.
data(tcga_kirc_example)

## Run the messinaSurv analysis on these data.  Use a tau
## objective, with a minimum performance of 0.6.  Note that
## messinaSurv analyses are very computationally-intensive,
## so multicore use with doMC loaded and parallel = TRUE is
## strongly recommended.  In this example we use a single
## core by default.
fit = messinaSurv(kirc.exprs, kirc.surv, obj_func = "tau", obj_min = 0.6)

## Plot the three best features found by Messina
plot(fit, indices = 1:3)

## Plot the best feature found by Messina, with 90% confidence bands.
## Note that the bootstrap iterations can be slow, so it is
## recommended that multiple cores are used, with doMC loaded
## and parallel = TRUE.
plot(fit, indices = 1, bootstrap_type = "ci", bootstrap_ci = 0.9)

## Plot the Messina fit of the 10th feature in the dataset, with
## +/- 1 standard deviation bands.
plot(fit, indices = 10, sort_features = FALSE, bootstrap_type = "stdev")

messina documentation built on Nov. 17, 2017, 9:12 a.m.