Skip to contents

Canproj: Cancer Projections

Source: R/canproj.R
canproj.Rd

Cancer incidence and mortality projections using the Canproj method. In brief, Canproj uses a decision-tree approach to determine the optimal model to project cancer data into the future based on historical trends. More information about this approach can be found here: https://doi.org/10.24095/hpcdp.40.9.02.

get_projections() extracts projection results from a canproj() object.

Usage

canproj(
  cdat,
  pdat,
  startp,
  standpop,
  projfor = "incidence",
  nagg = NULL,
  ncase = NULL,
  startage = NULL,
  newcohort = FALSE,
  ave5 = FALSE,
  sum5 = TRUE,
  methods = NULL,
  linkfunc = "power5",
  cuttrd = 0.04,
  shortp = 0,
  pD = 0.05,
  pGOF = 0.05
)

# S3 method for class 'canproj'
get_projections(object, ..., standpop = NULL)

# S3 method for class 'canproj'
print(x, ...)

Arguments

cdat

(age groups) * N (years) historical cancer data, 15<=N<=125.

pdat

(age groups) * (N + M) (years) observed and projected population, 5<=M<=25.

startp

The start calendar year of projection (e.g., 2009).

standpop

A StandardPopulation object that provides the weights (proportions) for each age group in a standard population.

projfor

Specify "incidence" or "mortality" if you want ASR as a criteria for nagg.

nagg

Number of years for data aggregation (by years). Default is 1-annual data.

ncase

Minimum number of cancer cases/deaths per year for splitting data.

startage

Youngest age group to include in the GLM. Default (NULL) picks based on mean cases.

newcohort

Assign new cohort effect as 0 (FALSE) or the last estimated cohort effect (TRUE), default is 0, use TRUE only if having evidence on negative new cohort effect.

ave5

ave5 = TRUE invokes the 5-year average method when age-only model is selected.

sum5

When the 5-year average method is used, sum5 = TRUE call the 5-year period based rate, otherwise (sum5 = FALSE), average the 5 rates in the 5 years for each age group.

methods

User required projection method can be specified by ADPC models: "nordpred" or "adpc-nb"; age-cohort models: "ac", "ac-nb"; age-period models: "age-trd", "com-trd"; and average: "ave5".

linkfunc

Link function. Default is "power5". Can be one of "log", "sqrt", or "identity".

cuttrd

Degenerating percent of trends per year after 5 years (shortp = 0) or the first projection year.

shortp

Attenuation percent of drift term or slope for the first 5 projection years.

pD

Trend selecting criteria of p-value of drift (linear trend) term.

pGOF

Model selection criteria of p-value of goodness-of-fit.

object

Output object from canproj().

...

Not currently used.

x

A object of class "canproj" to print.

Value

canproj() returns a named-list with class "canproj". The list contains the following:

  • annproj: A matrix of age-standardized rates and case counts.

  • agsproj: A data.frame of case counts for each age group and year.

  • method: The chosen method for projection. One of "ave5", "nordpred", "adpc-nb", "ac-poi", "ac-nb", "age-trd-nb", "age-trd-poi", "age-only", or "com-trd".

  • out: The output from function call for the chosen method (e.g., acproj(), adpcproj(), ave5proj(), hybdproj()).

  • obsy: Number of observed years of cancer data from cdat.

  • projy: Number of projected years based on pdat.

  • pdPC: A vector of p-values for the drift, period, and cohort effects in the "adpc" model.

get_projections() returns a data.frame.