Age-Period-Cohort Analysis
This plot shows heat map of the sparsity of a data matrix.
This plot shows sums of data matrix by age, period or cohort.
This plot shows time series of matrix within age, period or cohort.
Arrange data as an apc.data.list
Cut age, period and cohort groups from data set.
Computes age, period and cohort sums of a matrix
Fits an age period cohort model
Forecast for responses model with AC or CL structure.
Forecast for Poisson response model with AP structure.
Forecast models with APC structure.
Forecasts from age-period-cohort models.
Create design matrices
Get indices for mapping data into trapezoid formation
Imposing hypotheses on age-period-cohort models.
Identification of time effects
Implements direct tests between APC models
Estimate a single APC model
Generate table to select APC submodel
Make all descriptive plots.
Level plot of data matrix.
Make all fit plots.
Plot probability transform of responses given fitted values
Plots of apc estimates
Level plots of residuals / fitted values / linear predictors
Add connected line and standard deviation polygons to a plot
Age-period-cohort analysis
UK aids data
Asbestos data
Belgian lung cancer data
Italian bladder cancer data
Japanese breast cancer data
Motor data
Motor data
Motor data
US Casualty data, XL Group
2-sample mortality data.
Japanese breast cancer data
Internal apc Functions
Identification of time effects
Plots of apc estimates
Triangular matrices used in reserving
Functions for age-period-cohort analysis. Aggregate data can be organised in matrices indexed by age-cohort, age-period or cohort-period. The data can include dose and response or just doses. The statistical model is a generalized linear model (GLM) allowing for 3,2,1 or 0 of the age-period-cohort factors. Individual-level data should have a row for each individual and columns for each of age, period, and cohort. The statistical model for repeated cross-section is a generalized linear model. The statistical model for panel data is ordinary least squares. The canonical parametrisation of Kuang, Nielsen and Nielsen (2008) <DOI:10.1093/biomet/asn026> is used. Thus, the analysis does not rely on ad hoc identification.