confint {arules} | R Documentation |

Defines a method to compute confidence intervals for interest measures for association rules.

```
## S3 method for class 'rules'
confint(
object,
parm = "oddsRatio",
level = 0.95,
measure = NULL,
side = c("two.sided", "lower", "upper"),
method = NULL,
replications = 1000,
smoothCounts = 0,
transactions = NULL,
...
)
```

`object` |
an object of class rules. |

`parm` , `measure` |
name of the interest measures (see |

`level` |
the confidence level required. |

`side` |
Should a two-sided confidence interval or a one-sided limit be returned? Lower returns an interval with only a lower limit and upper returns an interval with only an upper limit. |

`method` |
method to construct the confidence interval. The available methods depends on the measure and the most common method is used by default. |

`replications` |
number of replications for method |

`smoothCounts` |
pseudo count for addaptive smoothing (Laplace smoothing). Often a pseudo counts of .5 is used for smoothing (see Detail Section). |

`transactions` |
if the rules object does not contain sufficient quality information, then a set of transactions to calculate the confidence interval for can be specified. |

`...` |
Additional parameters are ignored with a warning. |

This method creates a contingency table for each rule and then constructs a confidence interval for the specified measures.

Fast confidence interval approximations are currently available for the
measures `"support"`

, `"count"`

, `"confidence"`

, `"lift"`

, `"oddsRatio"`

, and `"phi"`

.
For all other measures, bootstrap sampling from a multinomial distribution
is used.

Haldan-Anscombe correction (Haldan, 1940; Anscombe, 1956) to avoids issues
with zero counts can be specified by `smoothCounts = 0.5`

. Here .5 is
added to each count in the contingency table.

Returns a matrix with with one row for each rule and the two columns
named `"LL"`

and `"UL"`

with the interval boundaries.
The matrix has the following additional attributes:

`measure` |
the interest measure. |

`level` |
the confidence level |

`side` |
the confidence level |

`smoothCounts` |
used count smoothing. |

`method` |
name of the method to create the interval |

`desc` |
description of the used method to calculate the confidence interval. The mentioned references can be found below. |

Michael Hahsler

Wilson, E. B. (1927). "Probable inference, the law of
succession, and statistical inference".
*Journal of the American Statistical Association,* 22 (158): 209-212.
doi:10.1080/01621459.1927.10502953

Clopper, C.; Pearson, E. S. (1934). "The use of confidence or fiducial
limits illustrated in the case of the binomial". *Biometrika,* 26 (4):
404-413.
doi:10.1093/biomet/26.4.404

Doob, J. L. (1935). "The Limiting Distributions of Certain Statistics".
*Annals of Mathematical Statistics,* 6: 160-169.
doi:10.1214/aoms/1177732594

Fisher, R.A. (1962). "Confidence limits for a cross-product ratio".
*Australian Journal of Statistics,* 4, 41.

Woolf, B. (1955). "On estimating the relation between blood group and
diseases". *Annals of Human Genetics,* 19, 251-253.

Haldane, J.B.S. (1940). "The mean and variance of the moments of chi-squared
when used as a test of homogeneity, when expectations are small".
*Biometrika,* 29, 133-134.

Anscombe, F.J. (1956). "On estimating binomial response relations".
*Biometrika,* 43, 461-464.

Other interest measures:
`coverage()`

,
`interestMeasure()`

,
`is.redundant()`

,
`is.significant()`

,
`support()`

```
data("Income")
# mine some rules with the consequent "language in home=english"
rules <- apriori(Income, parameter = list(support = 0.5),
appearance = list(rhs = "language in home=english"))
# calculate the confidence interval for the rules' odds ratios.
# note that we use Haldane-Anscombe correction (with smoothCounts = .5)
# to avoid issues with 0 counts in the contingency table.
ci <- confint(rules, "oddsRatio", smoothCounts = .5)
ci
# We add the odds ratio (with Haldane-Anscombe correction)
# and the confidence intervals to the quality slot of the rules.
quality(rules) <- cbind(
quality(rules),
oddsRatio = interestMeasure(rules, "oddsRatio", smoothCounts = .5),
oddsRatio = ci)
rules <- sort(rules, by = "oddsRatio")
inspect(rules)
# use confidence intervals for lift to find rules with a lift significantly larger then 1.
# We set the confidence level to 95%, create a one-sided interval and check
# if the interval does not cover 1 (i.e., the lower limit is larger than 1).
ci <- confint(rules, "lift", level = 0.95, side = "lower")
ci
inspect(rules[ci[, "LL"] > 1])
```

[Package *arules* version 1.7-7 Index]