sea.classbybreak {ldsa} | R Documentation |

## Collect Descriptive Statistics for Entailment Models as a Function of Breakpoint

### Description

Performs an enumeration of all response patterns associated with a series of error thresholds (as computed by `sea.breakpoints`

), and returns a number of descriptive statistics for each. This can be useful as an initial step in identifying candidates for confirmatory latent class analysis.

### Usage

sea.classbybreak(x, measure = c("rate", "z-score", "binom.p"),
max.classes = Inf, sea.table.precomp = NULL,
constructive = TRUE, ...)

### Arguments

`x` |
a `data.frame` with observations on rows; missing data is permitted. |

`measure` |
error measure to use; `rate` for absolute error rates, `z-score` for z-scores, or `binom.p` for p-values under an exact Binomial test. |

`max.classes` |
stop processing after encountering a threshold which generates at least `max.classes` legal response patterns. |

`sea.table.precomp` |
an object of class `sea.table` , computed on `x` (optional). |

`constructive` |
logical; should `sea.enumerate.constructive` be used to perform the enumeration? |

`...` |
additional arguments to `sea.enumerate` . |

### Details

`sea.classbybreak`

calls `sea.breakpoints`

to identify all error thresholds (based on `measure`

) at which new entailments are identified. It then proceeds to enumerate the set of legal response patterns for each threshold (using `sea.enumerate`

), retaining in each case the number of patterns (if any) and the fraction of `x`

which is covered by the pattern set. This process continues until all breakpoints have been explored, or until a threshold is identified for which the number of legal patterns is greater than or equal to `max.classes`

, at which point the algorithm terminates.

`sea.classbybreak`

has a plot method, which can be useful for identifying thresholds with favorable size/error ratios. These can then be examined in greater detail, e.g. using `sea.bclca.modelselect`

.

### Value

A three-column matrix of class `sea.classbybreak`

, whose columns contain (respectively):

`ErrorThresh ` |
Error threshold used |

`NumClasses ` |
Number of legal response patterns at the specified error threshold |

`Coverage ` |
The fraction of rows of `x` which are covered by the set of legal response patterns |

### Author(s)

Carter T. Butts buttsc@uci.edu

### See Also

`sea.breakpoints`

, `sea.entailment`

, `sea.enumerate`

, `sea.bclca.modelselect`

[Package

*ldsa* version 0.1-2

Index]