summary.pyears {survival}R Documentation

Summary function for pyears objecs


Create a printable table of a person-years result.


## S3 method for class 'pyears'
summary(object, header = TRUE, call = header, n = TRUE,
event = TRUE, pyears = TRUE, expected = TRUE, rate = FALSE, rr =expected,
ci.r = FALSE, ci.rr = FALSE, totals=FALSE, legend = TRUE, vline = FALSE,
vertical= TRUE, nastring=".", conf.level = 0.95,
scale = 1, ...)



a pyears object


print out a header giving the total number of observations, events, person-years, and total time (if any) omitted from the table


print out a copy of the call

n, event, pyears, expected

logical arguments: should these elements be printed in the table?

rate, ci.r

logical arguments: should the incidence rate and/or its confidence interval be given in the table?

rr, ci.rr

logical arguments: should the hazard ratio and/or its confidence interval be given in the table?


should row and column totals be added?


should a legend be included in the printout?


should vertical lines be included in the printed tables?


when there is only a single predictor, should the table be printed with the predictor on the left (vertical=TRUE) or across the top (vertical=FALSE)?


what to use for missing values in the table. Some of these are structural, e.g., risk ratios for a cell with no follow-up time.


confidence level for any confidence intervals


a scaling factor for printed rates


optional arguments which will be passed to the format function; common choices would be digits=2 or nsmall=1.


The pyears function is often used to create initial descriptions of a survival or time-to-event variable; the type of material that is often found in “table 1” of a paper. The summary routine prints this information out using one of pandoc table styles. A primary reason for choosing this style is that Rstudio is then able to automatically render the results in multiple formats: html, rtf, latex, etc.

If the pyears call has only a single covariate then the table will have that covariate as one margin and the statistics of interest as the other. If the pyears call has two predictors then those two predictors are used as margins of the table, while each cell of the table contains the statistics of interest as multiple rows within the cell. If there are more than two predictors then multiple tables are produced, in the same order as the standard R printout for an array.

The "N" entry of a pyears object is the number of observations which contributed to a particular cell. When the original call includes tcut objects then a single observation may contribute to multiple cells.


a copy of the object


The pandoc system has four table types: with or without vertical bars, and with single or multiple rows of data in each cell. This routine produces all 4 styles depending on options, but currently not all of them are recognized by the Rstudio-pandoc pipeline. (And we don't yet see why.)


Terry Therneau and Elizabeth Atkinson

See Also

cipoisson, pyears, format

[Package survival version 3.7-0 Index]