## Introduction

This vignette is not written yet! That is, it is under construction. With each new version of **eha**, this vignette will (hopefully) be more complete.

## What is **eha** in relation to **survival**?

The **eha** package can be seen as a complement to the recommended package **survival**: In fact, **eha** *depends* on **survival**, and for *standard Cox regression*, `eha::coxreg()`

simply calls `survival::agreg.fit()`

or `survival::coxph.fit()`

, functions *exported* by **survival**. The simple reason for this is that the underlying code in these **survival** functions is very fast and efficient. However, `eha::coxreg()`

has some unique features: *Sampling of risk sets*, *The “weird” bootstrap*, and *discrete time modeling* via maximum likelihood.

I have also put effort in producing nice and relevant printouts of regression results, both on screen and to \(\LaTeX\) documents (HTML output is on the TODO list). By *relevant* output I basically mean *avoiding misleading p-values*, show all *factor levels*, and use the *likelihood ratio test* in front of the *Wald test* where possible.

## Parametric survival models in **eha**

There is a special vignette describing the theory and implementation of the parametric failure time models. It is *not* very useful as a *user’s manual*.

### Accelerated Failure Time models

The parametric accelerated failure time (AFT) models are present via `eha::aftreg()`

, which is corresponding to `survival::survreg()`

. An important difference is that `eha::aftreg()`

allows for *left truncated data*.

### Parametric proportional hazards models

Parametric proportional hazards (PH) modeling is available through the functions `eha::phreg()`

and `eha::weibreg()`

, the latter still in the package for historical reasons. It will eventually be removed, since the Weibull distribution is also available in `eha::phreg()`

.

## Utilities

The primary applications in mind for **eha** were *demography* and *epidemiology*. There are some functions in **eha** that makes certain common tasks in that context easy to perform, for instance *rectangular cuts* in the *Lexis diagram*, creating *period* and *cohort* statistics, etc.