Main Content

Empirical cumulative distribution function

`[`

specifies additional options using one or more name-value arguments. For example,
`f`

,`x`

]
= ecdf(`y`

,`Name,Value`

)

specifies the type of
function for `'Function'`

,'survivor'`f`

as a survivor function.

`ecdf(___)`

produces a stairstep graph of
the evaluated function. The function visualizes interval estimates for interval-censored
data using shaded rectangles. You can specify

to include the confidence bounds
in the graph for fully observed, left-censored, right-censored, and double-censored
data.`'Bounds'`

,'on'

`ecdf`

computes the function values (`f`

) and the
confidence bounds (`flo`

and `fup`

) using different
algorithms, depending on the censorship information. The function type of
`f`

can be the cdf (default), Survivor Function, or Cumulative Hazard Function, as specified by the
`Function`

name-value argument.

Censorship Type | Algorithm for `f` | Algorithm for `flo` and `fup` |
---|---|---|

Right-censored data, which contains fully observed or right-censored observations |
Use the Kaplan-Meier estimator for the cdf and survivor function values. The Kaplan-Meier estimator $$\widehat{S}(t)$$ is given by $$\widehat{S}\left(t\right)={\displaystyle \prod _{{t}_{i}<t}\frac{{r}_{i}-{d}_{i}}{{r}_{i}}},$$ where *r*_{i}is the number of observations at risk at time*t*_{i}, and*d*_{i}is the number of failures at time*t*_{i}. For more details, see Kaplan-Meier Method.Use the Nelson-Aalen estimator for the cumulative hazard function values. The Nelson-Aalen estimator is given by $$\widehat{H}\left(t\right)={\displaystyle \sum _{{t}_{i}<t}\frac{{d}_{i}}{{r}_{i}}}.$$
| Use Greenwood’s formula, which is an approximation for the variance of the Kaplan-Meier estimator. The variance estimate is given by $$V\left(\widehat{S}\left(t\right)\right)={\widehat{S}}^{2}\left(t\right){\displaystyle \sum _{{t}_{i}<t}\frac{{d}_{i}}{{r}_{i}\left({r}_{i}-{d}_{i}\right)}}.$$ |

Left-censored data, which contains fully observed or left-censored observations | Use the Kaplan-Meier estimator. | Use Greenwood's formula. |

Double-censored data, which includes both right-censored and left-censored observations | Use Turnbull's algorithm [3][4]. You can specify
the maximum number of iterations ( | Use the Fisher information matrix. |

Interval-censored data, which includes interval-censored observations |
Use the expectation-maximization iterative convex minorant (EMICM) algorithm [5]. The EMICM algorithm uses either the EM algorithm or the ICM algorithm at each iteration. The `ICMFrequency` name-value argument determines the frequency of the ICM algorithm.`ecdf` runs the ICM step every specified number of iterations. By default,`ecdf` iterates the EM step nine times, runs the ICM step once, and then goes back to the EM step. You can specify the maximum number of iterations (`IterationLimit` ) and the termination tolerance on the function value (`Tolerance` ) for the algorithm.`ecdf` constructs mutually disjoint intervals, called Turnbull intervals, from the two-column matrix data`y` , and returns the Turnbull intervals (`x` ) and the estimates (`f` ) at the intervals. The left bounds of the intervals are from the first column of`y` , and the right bounds of the intervals are from the second column of`y` . For a fully observed observation (for the row with two of the same values`[t t]` ), the function converts`[t t]` to`[t–eps(t) t]` to create an interval with nonzero length before constructing the Turnbull intervals.
| Not supported |

[1] Cox, D. R., and D.
Oakes. *Analysis of Survival
Data*. London: Chapman & Hall,
1984.

[2] Lawless, J. F. *Statistical
Models and Methods for Lifetime Data*.
2nd ed., Hoboken, NJ: John Wiley & Sons, Inc.,
2003.

[3] Klein, John P., and Melvin L.
Moeschberger. *Survival Analysis: Techniques for Censored and Truncated
Data.* 2nd ed. Statistics for Biology and Health. New York: Springer,
2003.

[4] Turnbull, Bruce W. "Nonparametric
Estimation of a Survivorship Function with Doubly Censored Data." *Journal of the
American Statistical Association* 69, No. 345 (1974): 169–73.

[5] Anderson-Bergman, Clifford. "An
Efficient Implementation of the EMICM Algorithm for the Interval Censored NPMLE."
*Journal of Computational and Graphical Statistics* 26, no. 2 (April 3,
2017): 463–67.

[6] Ware, James H., and David L.
Demets. "Reanalysis of Some Baboon Descent Data." *Biometrics* 32, no. 2
(June 1976): 459–63.