getTestKM {Coxmos}

getTestKM

Description

This function computes and visualizes the Kaplan-Meier survival curve for a given test dataset, utilizing the cutoff derived from the original model. The function offers flexibility in terms of the type of Kaplan-Meier estimation, whether it's based on the linear predictor, PLS components, or original variables.

Usage

getTestKM(
  model,
  X_test,
  Y_test,
  cutoff,
  type = "LP",
  ori_data = TRUE,
  BREAKTIME = NULL,
  n.breaks = 20,
  title = NULL
)

Arguments

model

Coxmos model.

X_test

Numeric matrix or data.frame. Explanatory variables for test data (raw format). Qualitative variables must be transform into binary variables.

Y_test

Numeric matrix or data.frame. Response variables for test data. Object must have two columns named as "time" and "event". For event column, accepted values are: 0/1 or FALSE/TRUE for censored and event observations.

cutoff

Numeric. Cutoff value to split the observations into two groups. Recommended to compute optimal cutoff value with getAutoKM() function.

type

Character. Kaplan Meier for complete model linear predictor ("LP"), for PLS components ("COMP") or for original variables ("VAR") (default: LP).

ori_data

Logical. Compute the Kaplan-Meier plot with the raw-data or the normalize-data to compute the best cut-point for splitting the data into two groups. Only used when type = "VAR" (default: TRUE).

BREAKTIME

Numeric. Size of time to split the data into "total_time / BREAKTIME + 1" points. If BREAKTIME = NULL, "n.breaks" is used (default: NULL).

n.breaks

Numeric. If BREAKTIME is NULL, "n.breaks" is the number of time-break points to compute (default: 20).

title

Character. Kaplan-Meier plot title (default: NULL).

Details

The getTestKM function is designed to evaluate the survival probabilities of a test dataset based on a pre-trained Coxmos model. The function ensures that the test times are consistent with the training times. Depending on the specified type, the function can compute the Kaplan-Meier curve using:

• The complete model's linear predictor (LP).

• The PLS components (COMP).

• The original variables (VAR).

For the LP type, the function predicts scores for the X_test and subsequently predicts the linear predictor using these scores. For the COMP type, the function predicts scores for each component in the model and computes the Kaplan-Meier curve for each. For the VAR type, the function computes the Kaplan-Meier curve for each variable in the test dataset.

The function also provides the flexibility to compute the Kaplan-Meier plot using raw data or normalized data, which can be useful for determining the optimal cut-point for data segmentation. The time intervals for the Kaplan-Meier estimation can be defined using either the BREAKTIME or n.breaks parameters.

The resulting Kaplan-Meier plot provides a visual representation of the survival probabilities over time, segmented based on the specified cutoff. This allows for a comprehensive evaluation of the test dataset's survival characteristics in the context of the original model.

Value

Depending on the specified type parameter, the function returns:

• LP: A ggplot object visualizing the Kaplan-Meier survival curve based on the linear predictor, segmented by the specified cutoff.

• COMP: A list of ggplot objects, where each plot represents the Kaplan-Meier survival curve for a specific PLS component in the model, segmented by the respective cutoffs.

• VAR: A list of ggplot objects, where each plot visualizes the Kaplan-Meier survival curve for a specific variable in the test dataset, segmented by the respective cutoffs.

Each plot provides a visual representation of the survival probabilities over time, allowing for a comprehensive evaluation of the test dataset's survival characteristics in the context of the original model.

Author(s)

Pedro Salguero Garcia. Maintainer: [e-mail]

References

Kaplan EL, Kaplan EL, Meier P (1958). “Nonparametric Estimation from Incomplete Observations.” Journal of the American Statistical Association. doi:10.1007/978-1-4612-4380-9_25, https://link.springer.com/chapter/10.1007/978-1-4612-4380-9_25.

Examples

data("X_proteomic")
data("Y_proteomic")
set.seed(123)
index_train <- caret::createDataPartition(Y_proteomic$event, p = .5, list = FALSE, times = 1)
X_train <- X_proteomic[index_train,1:50]
Y_train <- Y_proteomic[index_train,]
X_test <- X_proteomic[-index_train,1:50]
Y_test <- Y_proteomic[-index_train,]
splsicox.model <- splsicox(X_train, Y_train, n.comp = 2, penalty = 0.5, x.center = TRUE,
x.scale = TRUE)
KMresult = getAutoKM(type = "LP", model = splsicox.model)
cutoff <- getCutoffAutoKM(result = KMresult)
getTestKM(splsicox.model, X_test, Y_test, cutoff)