avg: Model averaging

Description Usage Arguments Details Value Author(s) References Examples

Description

Average model estimates according to Akaike's Information Criterion.

Usage

1
2

Arguments

...

Possibly named objects of class "ea" or "drm".

dose

Doses for which model-averaged responses need to be obtained.

Details

Starting from a set of K plausible candidate models, the averaged model estimate \hat{θ} is defined as ∑_{i=1}^{K}{w_i θ_i}. The weights are defined as

w_i = \frac{exp(-0.5 Δ_i)}{∑_{j=1}^{K}{exp(-0.5 Δ_j)}}

with Δ_i = AIC_i - AIC_{min}

The variance of the averaged model estimate is given by

var(\hat{θ}) = [∑_{i=1}^{K}{w_i √{var(θ_i) + (θ_i - \hat{θ})^2}}]^2

Value

An object of S3 class "avg".

Author(s)

[email protected]

References

Examples

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
## Exposure assessment from concentration data
gam <- ea_conc(x = x, d = d, data = giardia, model = "gamma")
lno <- ea_conc(x = x, d = d, data = giardia, model = "lognormal")
wei <- ea_conc(x = x, d = d, data = giardia, model = "weibull")
inv <- ea_conc(x = x, d = d, data = giardia, model = "invgauss")

## Model averaging
avg_ea("Gamma" = gam,
       "Log-Normal" = lno,
       "Weibull" = wei,
       "Inverse Gaussian" = inv)


## Fit several dose-response models to the Campylobacter dataset
bp <- drm(infected, total, dose, campy, "betapoisson")
ll <- drm(infected, total, dose, campy, "loglogistic")
lp <- drm(infected, total, dose, campy, "logprobit")
ev <- drm(infected, total, dose, campy, "extremevalue")

## Model averaging
avg_drm("bp" = bp, "ll" = ll, "lp" = lp, "ev" = ev,
        dose = c(1, 10, 100))

brechtdv/QMRA documentation built on May 13, 2019, 5:06 a.m.