fullsim: Simulated artificial data with covariates in both the gating...
In senhu/mvClaim: Multivariate general insurance claims modelling

Description Usage Format Details Source

An artificial data set of size 500 containing the bivariate response y1, y2, three covariates w1, w2, w3, and the true labels.

1
2
3

fullsim

fullsim.test

A data frame with 500 rows and 6 variables:

y1,y2: The bivariate response variable.
w1,w2,w3: Three covariates.
label: Their true labels.

This data set of size 500 is artificially generated as following: first w1, w2, and w3 are simulated from Normal(0, 0.3). The true number of component is G = 2. For the first component, the parameters are generated as:

α1 = exp(1 + 0.2 w1 + 0.2 w2)

α2 = exp(0.1 + 0.1 w2 + 0.1 w3)

α3 = exp(0.5 + 0.2 w1 + 0.2 w2 + 0.2 w3)

β = exp(0.2 + 0.1 w1 + 0.1 w2 + 0.2 w3)

For the second component, the parameters are generated as:

α1 = exp(0.1 + 0.1 w1 + 0.1 w2)

α2 = exp(2 + 0.3 w2 + 0.3 w3)

α3 = exp(1.5 + 0.2 w1 + 0.1 w2 + 0.1 w3)

β = exp(0.7 + 0.1 w1 + 0.1 w2 + 0.2 w3)

For each set of parameters (α1, α2, α3, β) a random sample from this bivariate gamma distribution is simulated, i.e. y ~ BG(α1, α2, α3, β). The gating is simulated from a logistic regression model where the regression coeffcients are

logit(p) = 10 + 40 w1 + 30 w2 + 100 w3 ,

based on which a binomial simulation is used, and the two components are sampled from the simulations above to form the data set. The final simulated data set consists of 232 samples from component 1 and 268 observations from component 2.

Hu, S., Murphy, T. B. and O'Hagan, A. (2019) Bivariate Gamma Mixture of Experts Models for Joint Claims Modeling. To appear.

senhu/mvClaim documentation built on Jan. 29, 2022, 3:18 p.m.