EMICM {Icens} | R Documentation |

## Compute the NPMLE for censored data using the EMICM.

### Description

An implementation of the hybrid EM ICM (Iterative convex minorant)
estimator of the distribution function proposed by Wellner and Zahn (1997).

### Usage

EMICM(A, EMstep=TRUE, ICMstep=TRUE, keepiter=FALSE, tol=1e-07,
maxiter=1000)

### Arguments

`A` |
Either the m by n clique matrix or the n by 2 matrix
containing the event time intervals. |

`EMstep` |
Boolean, indicating whether to take an EM step in the
iteration. |

`ICMstep` |
Boolean, indicating whether to take an ICM step. |

`keepiter` |
Boolean determining whether to keep the iteration
states. |

`tol` |
The maximal L1 distance between successive estimates
before stopping iteration. |

`maxiter` |
The maximal number of iterations to perform before
stopping. |

### Details

Lots, and they're complicated too!

### Value

An object of class `icsurv`

containing the following
components:

`pf ` |
The estimated probabilities. |

`sigma ` |
The NPMLE of the survival function on the maximal
antichains. |

`weights ` |
The diagonal of the likelihood function's second
derivative. |

`lastchange ` |
A vector of differences between the last two
iterations. |

`numiter ` |
The total number of iterations performed. |

`iter ` |
Is only present if `keepiter` is true; states of
sigma during the iteration. |

`intmap ` |
The real representation associated with the
probabilities reported in `pf` . |

### Author(s)

Alain Vandal and Robert Gentleman

### References

*A hybrid algorithm for computation of the nonparametric
maximum likelihood estimator from censored data*, J. A. Wellner and
Y. Zhan, 1997, JASA.

### See Also

`EM`

,`VEM`

, `PGM`

### Examples

data(cosmesis)
csub1 <- subset(cosmesis, subset=Trt==0, select=c(L,R))
EMICM(csub1)
data(pruitt)
EMICM(pruitt)

[Package

*Icens* version 0.5

Index]