# t ime, e vent(s) and n umber at risk. ```r ten(x, ...) ## S3 method for class 'numeric' ten(x, ...) ## S3 method for class 'Surv' ten(x, ..., call = NULL) ## S3 method for class 'coxph' ten(x, ..., abbNames = TRUE, contrasts.arg = NULL) ## S3 method for class 'survfit' ten(x, ..., abbNames = TRUE, contrasts.arg = NULL) ## S3 method for class 'formula' ten(x, ..., abbNames = TRUE, contrasts.arg = NULL) ## S3 method for class 'data.frame' ten(x, ..., abbNames = TRUE, contrasts.arg = NULL, call = NULL) ## S3 method for class 'data.table' ten(x, ..., abbNames = TRUE, mm = NULL, call = NULL) ## S3 method for class 'ten' ten(x, ..., abbNames = NULL, call = NULL) ``` ## Arguments - `x`: For the default method, a `numeric` vector indicating an **event** (or status). Each element indicates whether an event occurred (`1`) or not (`0`) for an observation. These are assumed to be ordered by discrete times. This is similar to the `event` argument for `Surv` objects. Methods are available for objects of class `Surv`, `survfit`, `coxph` and `formula`. - `...`: Additional arguments (not implemented). - `call`: Used to pass the `call` from a `formula` to the final `ten.data.table` method. - `abbNames`: Abb reviate names? If `abbNames="TRUE"` (the default), the covariate groups are referred to by number. As the names for each covariate group are made by concatenating the predictor names, the full names can become unwieldly. If `abbNames="FALSE"`, the full names are given. In either case, the `longNames` are given as an `attribute` of the returned `ten` object. - `contrasts.arg`: Methods for handling factors. A `list`. The `names` are the names of columns of the `model.frame` containing `factor`s. The **values** are used as replacement values for the `stats::contrasts` replacement function. These should be functions (given as character strings) or numeric matrices. This can be passed from `survfit`, `coxph` and `formula` objects to: ?stats::model.matrix - `mm`: Used to pass the `model.matrix` from a `formula` to the final `ten.data.table` method. ## Returns A `data.table` with the additional `class` `ten`. By default, the shape returned is 'long' i.e. there is one row for each unique timepoint per covariate group. The basic form, for a `numeric` or `Surv` object, has columns: - **t**: t ime. - **e**: number of e vents. - **n**: n umber at risk. A `survfit`, `coxph` or `formula` object will have additional columns: - **cg**: c ovariate g roup. This is formed by combining the variables; these are separated by a comma ','. - **ncg**: n umber at risk, by c ovariate g roup Special terms . The following are considered 'special' terms in a survival model: - **strata**: For a stratified model, `ten` returns a `list` with one element per strata, which is a `ten` object. This has the class `stratTen`. The name of the list elements are those of the strata in the model. - **cluster**: These terms are dropped. - **tt**: The variable is unchanged. That is, time-transform terms are handled as if the the function `tt(x)` was `identity(x)`. Attribures . The returned object will also have the following `attributes`: - **shape**: The default is `"long"` but is changed to `"wide"` when `asWide` is called on the object. - **abbNames**: Abbreviate names? - **longNames**: A `data.table` with two columns, showing the abbrevbiated and full names. - **ncg**: Number of covariate groups - **call**: The call used to generate the object - **mm**: The `model.matrix` used to generate to generate the object, if applicable. Additional attributes will be added by the following functions: `sf` `ci` ## Note The methods for `data.frame` (for a model frame) and `data.table` are not typically intended for interactive use. Currently only binary status and right-censoring are supported. In stratified models, only one level of stratification is supported (i.e. strata cannot be 'nested' currently). Partial matching is available for the following arguments, based on the characters in bold: * abb Names * con trasts.arg ## Examples ```r require("survival") ## binary vector ten(c(1, 0, 1, 0, 1)) ## Surv object df0 <- data.frame(t=c(1, 1, 2, 3, 5, 8, 13, 21), e=rep(c(0, 1), 4)) s1 <- with(df0, Surv(t, e, type="right")) ten(s1) ## some awkward values suppressWarnings( s1 <- Surv(time=c(Inf, -1, NaN, NA, 10, 12), event=c(c(NA, 1, 1, NaN, Inf, 0.75)))) ten(s1) ## coxph object ## K&M. Section 1.2. Table 1.1, page 2. data("hodg", package="KMsurv") hodg <- data.table::data.table(hodg) data.table::setnames(hodg, c(names(hodg)[!names(hodg) %in% c("score", "wtime")], "Z1", "Z2")) c1 <- coxph(Surv(time=time, event=delta) ~ Z1 + Z2, data=hodg[gtype==1 & dtype==1, ]) ten(c1) data("bmt", package="KMsurv") ten(c1 <- coxph(Surv(t2, d3) ~ z3*z10, data=bmt)) ## T&G. Section 3.2, pg 47. ## stratified model data("pbc", package="survival") c1 <- coxph(Surv(time, status==2) ~ log(bili) + age + strata(edema), data=pbc) ten(c1) ## K&M. Example 7.2, pg 210. data("kidney", package="KMsurv") with(kidney[kidney$type==2, ], ten(Surv(time=time, event=delta))) s1 <- survfit(Surv(time=time, event=delta) ~ type, data=kidney) ten(s1)[e > 0, ] ## A null model is passed to ten.Surv (t1 <- with(kidney, ten(Surv(time=time, event=delta) ~ 0))) ## but the original call is preserved attr(t1, "call") ## survival::survfit doesn't accept interaction terms... ## Not run: s1 <- survfit(Surv(t2, d3) ~ z3*z10, data=bmt) ## End(Not run) ## but ten.formula does: ten(Surv(time=t2, event=d3) ~ z3*z10, data=bmt) ## the same is true for the '.' (dot operator) in formulas (t1 <- ten(Surv(time=t2, event=d3) ~ ., data=bmt)) ## impractical long names stored as an attribute attr(t1, "longNames") ## not typically intended to be called directly mf1 <- stats::model.frame(Surv(time, status==2) ~ age + strata(edema) + strata(spiders), pbc, drop.unused.levels = TRUE) ten(mf1) ``` ## See Also `asWide` `print`

ten function