Fit Multiple Models

Source: R/api_fit_outlier.R

Conduct multiple novelty tests for a new observation

fit_multiple_models(
  A,
  z,
  response,
  alpha = 0.05,
  type = "fwd",
  q = 0.5,
  comp = NULL,
  nsim = 10000,
  ncores = 1,
  trace = TRUE,
  validate = TRUE
)

Arguments

A

A character matrix or data.frame
z

Named vector (same names as colnames(A) but without the class variable)
response

A character with the name of the class variable of interest
alpha

The significance level
type

Character ("fwd", "bwd", "tree" or "tfwd") - the type of interaction graph to be used
q

Penalty term in the stopping criterion when fitting the interaction graph (0 = AIC and 1 = BIC)
comp

A list with character vectors. Each element in the list is a component in the graph (using expert knowledge)
nsim

Number of simulations
ncores

Number of cores to use in parallelization
trace

Logical indicating whether or not to trace the procedure
validate

Logical. If true, it checks if A has only single character values and converts it if not.

Value

An object of type multiple_models; a list of of novely objects from which one can query pvalues etc. for outlierdetection.

See also

fit_outlier, fit_mixed_outlier

Examples


library(dplyr)
set.seed(1)

# A patient with psoriasis
z <- unlist(derma[2, 1:10])

d <- derma[, c(names(z), "ES")] %>%
     filter(ES %in% c("chronic dermatitis", "psoriasis"))

m <- fit_multiple_models(d, z, "ES", nsim = 1000, trace = FALSE, validate = FALSE)

plot(m)
#> Picking joint bandwidth of 0.619
print(m)
#> $psoriasis
#> 
#>  -------------------------------- 
#>   Simulations: 1000 
#>   Variables: 10 
#>   Observations: 112 
#>   Estimated mean: 13.79 
#>   Estimated variance: 10.17 
#>  --------------------------------
#>   Critical value: 19.44595 
#>   Deviance: 17.96901 
#>   P-value: 0.092 
#>   Alpha: 0.05 
#>   <novelty, outlier_model, list> 
#>  --------------------------------
#> 
#> $`chronic dermatitis`
#> 
#>  -------------------------------- 
#>   Simulations: 1000 
#>   Variables: 10 
#>   Observations: 49 
#>   Estimated mean: 10.61 
#>   Estimated variance: 9.64 
#>  --------------------------------
#>   Critical value: 18.83373 
#>   Deviance: 34.74245 
#>   P-value: 0 
#>   Alpha: 0.05 
#>   <novelty, outlier_model, list> 
#>  --------------------------------
#> 
#> attr(,"class")
#> [1] "multiple_models" "list"