format_iboost: Function to calculate selective inference based on sampled...
In davidruegamer/iboost: Inference for model-based boosting

Description Usage Arguments Details Examples

This function calculates p-values and confidence intervals based on sampled values and weights obtained by the iboost function.

1	format_iboost_res(res, alpha, method, vT, this_var, computeCI, fac = 2, y)

`res`	list of sampled values either directly obtained in the `iboost` call or when using `iboost` with `returnSamples = TRUE`.
`alpha`	numeric; level for confidence interval(s)
`method`	character; method used for sampling
`vT`	list of test vector(s)
`this_var`	numeric; variance to be used for inference
`computeCI`	logical; whether or not to compute confidence intervals
`fac`	numeric; used in inverse search for confidence interval limits
`y`	vector; response vector

The specific use of this function is to (internally) compute inference based on samples in the iboost function, to recalculate inference for given samples dist from an iboost object or to calculate inference in the first place when iboost has been ran with argument returnSamples = TRUE. In the first and third case, use format_iboost_res, otherwise format_iboost.

if(require("mboost")){

set.seed(0)

n <- 200
x1 <- rnorm(n)
x2 <- rnorm(n) + 0.25 * x1
x3 <- rnorm(n)
eta <- 3 * sin(x1) + x2^2
y <- scale(eta + rnorm(n), scale = FALSE)

spline1 <- bbs(x1, knots = 20, df = 4)
knots.x2 <- quantile(x2, c(0.25, 0.5, 0.75))
spline2 <- bbs(x2, knots = knots.x2, df = 4)
spline3 <- bbs(x3, knots = 20, df = 4)

data <- data.frame(y=y, x1=x1, x2=x2, x3=x3)

mod1 <- mboost(y ~ spline1 + spline2 + spline3,
control=boost_control(mstop = 73), offset = 0, 
data = data)

# calculate p-values and intervals for model with 
# fixed stopping iteration:
# this is done with only B = 100 samples for
# demonstrative purposes and should be increased 
# for actual research questions
res <- iboost(mod1, method = "impsamp", B = 100)

# recalculate inference for different variance or alpha level
format_iboost(res, alpha = 0.1, this_var = var(y)*5)
}