init (CoxPH)

  • Available in: CoxPH

  • Hyperparameter: no

Description

When building a CoxPH model, the init option specifies the initial value, \(\beta^{(0)}\), for the coefficient vector. This value defaults to 0.

Example

R

Python

library(h2o)
h2o.init()

# import the heart dataset
heart <- h2o.importFile("http://s3.amazonaws.com/h2o-public-test-data/smalldata/coxph_test/heart.csv")

# split the dataset into train and validation datasets
heart_split <- h2o.splitFrame(data = heart, ratios = 0.8, seed = 1234)
train <- heart_split[[1]]
test <- heart_split[[2]]

# train your model
coxph_model <- h2o.coxph(x = "age",
                         event_column = "event",
                         start_column = "start",
                         stop_column = "stop",
                         training_frame = heart,
                         init = 3)

# view the model details
coxph_model
Loading required namespace: survival
Model Details:
==============

H2OCoxPHModel: coxph
Model ID:  CoxPH_model_R_1570809926481_1
Call:
Surv(start, stop, event) ~ age

      coef exp(coef) se(coef)    z     p
age 0.0307    1.0312   0.0143 2.15 0.031

Likelihood ratio tes = 6109  on 1 df, p =< 2e-16
n = 172, number of events = 75

import h2o
from h2o.estimators.coxph import H2OCoxProportionalHazardsEstimator
h2o.init()

# import the heart dataset
heart = h2o.import_file("http://s3.amazonaws.com/h2o-public-test-data/smalldata/coxph_test/heart.csv")

# split the dataset into train and test datasets
train, test = heart.split_frame(ratios = [.8], seed=1234)

#specify the init parameter's value
init = 3

# initialize an train a CoxPH model
coxph = H2OCoxProportionalHazardsEstimator(start_column="start",
                                           stop_column="stop",
                                           ties="breslow",
                                           init=init)
coxph.train(x="age", y="event", training_frame=heart)

# view the model details
coxph.train
Model Details
=============
H2OCoxProportionalHazardsEstimator :  Cox Proportional Hazards
Model Key:  CoxPH_model_python_1570808496252_2

Call:
Surv(start, stop, event) ~ age

Coefficients: CoxPH Coefficients
names    coefficients    exp_coef    exp_neg_coef    se_coef    z_coef
-------  --------------  ----------  --------------  ---------  --------
age      0.030691        1.03117     0.969775        0.0142686  2.15095

Likelihood ratio test=5.160759
n=172, number of events=75

Scoring History:
    timestamp            duration    iterations    loglik
--  -------------------  ----------  ------------  --------
    2019-10-11 08:59:31  0.000 sec   0             -298.326
    2019-10-11 08:59:31  0.001 sec   1             -295.799
    2019-10-11 08:59:31  0.002 sec   2             -295.745
    2019-10-11 08:59:31  0.004 sec   3             -295.745
<bound method H2OEstimator.train of >