summary.aareg: Summarize an aareg fit

Description

Creates the overall test statistics for an Aalen additive regression model

Usage

# S3 method for aareg
summary(object, maxtime, test=c("aalen", "nrisk"), scale=1,...)

Arguments

object

the result of a call to the aareg function

maxtime

truncate the input to the model at time "maxtime"

test

the relative time weights that will be used to compute the test

scale

scales the coefficients. For some data sets, the coefficients of the Aalen model will be very small (10-4); this simply multiplies the printed values by a constant, say 1e6, to make the printout easier to read.

…

for future methods

Value

a list is returned with the following components

table

a matrix with rows for the intercept and each covariate, and columns giving a slope estimate, the test statistic, it's standard error, the z-score and a p-value

test

the time weighting used for computing the test statistics

test.statistic

the vector of test statistics

test.var

the model based variance matrix for the test statistic

test.var2

optionally, a robust variance matrix for the test statistic

chisq

the overall test (ignoring the intercept term) for significance of any variable

a vector containing the number of observations, the number of unique death times used in the computation, and the total number of unique death times

Details

It is not uncommon for the very right-hand tail of the plot to have large outlying values, particularly for the standard error. The maxtime parameter can then be used to truncate the range so as to avoid these. This gives an updated value for the test statistics, without refitting the model.

The slope is based on a weighted linear regression to the cumulative coefficient plot, and may be a useful measure of the overall size of the effect. For instance when two models include a common variable, "age" for instance, this may help to assess how much the fit changed due to the other variables, in leiu of overlaying the two plots. (Of course the plots are often highly non-linear, so it is only a rough substitute). The slope is not directly related to the test statistic, as the latter is invariant to any monotone transformation of time.

Examples

Run this code

# NOT RUN {
afit <- aareg(Surv(time, status) ~ age + sex + ph.ecog, data=lung,
     dfbeta=TRUE)
summary(afit)
# }
# NOT RUN {
              slope   test se(test) robust se     z        p 
Intercept  5.05e-03    1.9     1.54      1.55  1.23 0.219000
      age  4.01e-05  108.0   109.00    106.00  1.02 0.307000
      sex -3.16e-03  -19.5     5.90      5.95 -3.28 0.001030
  ph.ecog  3.01e-03   33.2     9.18      9.17  3.62 0.000299

Chisq=22.84 on 3 df, p=4.4e-05; test weights=aalen
# }
# NOT RUN {
summary(afit, maxtime=600)
# }
# NOT RUN {
              slope   test se(test) robust se      z        p 
Intercept  4.16e-03   2.13     1.48      1.47  1.450 0.146000
      age  2.82e-05  85.80   106.00    100.00  0.857 0.392000
      sex -2.54e-03 -20.60     5.61      5.63 -3.660 0.000256
  ph.ecog  2.47e-03  31.60     8.91      8.67  3.640 0.000271

Chisq=27.08 on 3 df, p=5.7e-06; test weights=aalen
# }