Using the age- and cause-specific mortality rates, this function applies hierarchical clustering to suggest possible cause-of-death groupings that isolate the causes that are suceptible to contribute to the young adult mortality hump.

1 |

`data` |
list produced with |

`x.range` |
age range to consider for the analysis |

`k` |
either a fixed number of cluster or the name of the criterion to use for selection |

This function is designed to help selecting the causes of death that contribute to the young adult mortality hump. It procedes in steps.

Compute the first derivative of the force of mortality (

`rxc`

) for each cause, in order to focuse on the amount of deviation instead of the absolute death rate.Using the provided age-range, compute the euclidian distance between each couple of causes.

Based on this distance, run a hierchical clustering method (

`"complete"`

algorithm of the hclust function).If

`k`

is numerical, it is taken as the chosen number of clusters and each cause of death is assigned to one of the k groups. Alternatively,`k`

can indicate one of the selection criteria available in the`WeightedCluster`

package. Among the most interesting options, is the Average Silhouette Width (`ASW`

) that compares the average distance of an observation from the other members of its group and its average distance from the closest group. The`ASW`

is computed for each number of groups`k`

, and the one maximising the`ASW`

is selected for its ability to maximise the homogeneity within the groups and the heterogeneity between the groups.

codgroup returns a list of six elements containing

- cluster
An object of class hclust on which additional analysis can be performed.

- groups
A membership vector indicating a group number for each of the causes of death.

- k
The number of groups chosen.

- typ
A list of groups of causes as needed by the function codhump.

- data
The original data stored in a list produced with

`HCD2MH`

or similarly structured- x.range
A vector indicating the age range to consider for the analysis

`codhump`

, `HCD2MH`

, `as.clustrange`

1 2 3 |

Questions? Problems? Suggestions? Tweet to @rdrrHQ or email at ian@mutexlabs.com.

All documentation is copyright its authors; we didn't write any of that.