data.aids: UK aids data

Description

Function that organises UK aids data in apc.data.list format.

The data set is taken from table 1 of De Angelis and Gilks (1994). The data are also analysed by Davison and Hinkley (1998, Example 7.4). The data are reporting delays for AIDS counting the number of cases by the date of diagnosis and length of reporting delay, measured by quarter.

The data set is in "trapezoid"-format. The original data set is unbalanced in various ways: first column covers a reporting delay of less than one month (or should it be less than one quarter?); last column covers a reporting delay of at least 14 quarters; last diagonal include incomplete counts. The default data set excludes the incomplete counts in the last diagonal, but includes the unbalanced first and last columns.

Usage

data.aids(all.age.groups = FALSE)

Arguments

all.age.groups

logical. If FALSE (default), the last calendar year with incomplete counts is ignored.

Value

The value is a list in apc.data.list format.

response

matrix of cases

data.format

logical equal to "trapezoid".

age1

numeric equal to 0. This is the label for the reporting delay.

per1

NULL. Not needed when data.format="trapezoid"

coh1

numeric equal to 1983.5. This is the label for the diagnosis quarter (1983, third quarter).

unit

numeric equal to 1/4. This is the width of the age and period groups.

per.zero

numeric equal to 0.

per.max

numeric equal to 38.

time.adjust

numric equal to 0.

label

character. Default data has "UK AIDS - clean".

References

De Angelis, D. and Gilks, W.R. (1994) Estimating acquired immune deficiency syndrome incidence accounting for reporting delay. Journal of the Royal Statistical Sociey A 157, 31-40.

Davison, A.C. and Hinkley, D.V. (1998) Bootstrap methods and their application. Cambridge: Cambridge University Press.

Examples

Run this code

# NOT RUN {
#########################
##	It is convient to construct a data variable
data	<- data.Belgian.lung.cancer()
##	To see the content of the data
data

#########################
#	Forecast AIDS incidences by diagonsis year (cohort).
#	uses as poisson response model with an AC structure
#	although there is evidence of overdispersion and the
#	period effect appears significant.
#	The omission of the period effect follows
#	Davison and Hinkley and a parsimoneous model may be
#	advantageous when forecasting.
#
apc.fit.table(data.aids(),"poisson.response")
fit <- apc.fit.model(data.aids(),"poisson.response","AC")
forecast <- apc.forecast.ac(fit)
data.sums.coh <- apc.data.sums(data.aids())$sums.coh
forecast.total <- forecast$response.forecast.coh
forecast.total[,1]	<- forecast.total[,1]+data.sums.coh[25:38]
x	<- seq(1983.5,1992.75,by=1/4)
y	<- data.sums.coh
xlab<- "diagnosis year (cohort)"
ylab<- "diagnoses"
main<- "Davison and Hinkley, Fig 7.6, parametric version"
plot(x,y,xlim=c(1988,1993),ylim=c(200,600),xlab=xlab,ylab=ylab,main=main)
apc.polygon(forecast.total,x.origin=1989.25,unit=1/4)
# }