In bayesiandemography/demarray: Demographic Arrays

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)

library(demarray)

Introduction

op

x is any array with complete metadata

| Condition | y is Counts | y is Values | y is array | |:-----------------------------|:--------------|:-------- -----|:-------------| | x has dimensions not present in y | Throw errror | Add dimensions to y | Throw error | | y has dimensions not present in x | Collapse dimensions in y | Throw | Throw error | | On shared dimension, categories for x have many-to-one relationship to categories for y | Throw error | Split categories in y to match x | Throw error | | On shared dimension, categories for y have many-to-one relationship to categories for y | Collapse categories in y to match x | Throw error | Throw error |

Labels

Default

| dimlabels | example | |:----------------|:---------------------| | Categories | "New Zealand", "Australia" |

• only restriction is that unique and non-blank (NA allowed)
• iterations are categories, convention that use 1, 2, 3

Integers

"2020", "2021", "2022" "<0", "5", "10", "15+" Cannot have duplicates. Can have open, NA

Quantities

Plain integers "2020", "5", "0"

Groups "1-4", "5-9", "100-999"

Open "<2000", "100+"

Mixtures "0", "1-4", "5-9", "10+"

Not necessarily covering whole range "0-4", "5-9", "15-19", "20-24"

Intervals

Intervals deliberately ambiguous more ambiguous than DateRanges. Conventional labels for years are ambiguous. Using intervals, can represent info in labels without forcing user to clarify, except when essential to do so.

"<2000", "2000-2005", "2005-2010", "2010-2015", "2015-2010", 2020+", "2000-2001"

Calendar

| dimlabel | example | |:----------------|:---------------------| | Quarters | "<1990 Q4", 2000 Q1", "2020 Q3+" | | Months | "<2000 Jan", "2020 Aug", "2025 Feb+" | | DatePoints | "2020-01-01", "2005-06-01" | | DateRanges | "[2020-01-01, 2020-12-31]", (-Inf, 2019-12-31]" |

Specialised

| dimlabels | dimtype | example | |:--------------|:----------|:--------------| | Triangles | triangles | "Lower", "Upper" | | Directions | directions | "In", "Out" | | Quantiles | quantiles | "2.5%", "50%", "97.5%" |

• if 'strict' is TRUE, dimensions with the indicated dimtype must have the associated dimlabels
• can use these dimlabels with other dimtypes

Notes

Messy - aim is to represent messy labeling conventions - inherits messiness from the conventions - tangled relationship - everything a special case of Categories - original labeling conventions ambiguous. Deliberately carry that ambiguity over to the labels - "lazy disambiguation" - only disambiguate when we need to

rather than impose new unambiguous labelling system, allow people to work with existing ambiguous system

bayesiandemography/demarray documentation built on Dec. 28, 2021, 8:46 a.m.