plotSurface1d: Plot posterior predictive exposure response surface for an...

In jantonelli111/NLinteraction: Bayesian variable selection in multivariate semi-parametric regression models

Description

This function takes in an object which is a model from the NLint function and plots the exposure response curve and pointwise 95 band for exposure j1, while fixing exposure j2 at quantile_j2, the remaining exposures at quantile_rest, and the covariates at their mean

Usage

plotSurface1d(
  NLmod,
  X,
  C,
  j1,
  j2,
  gridLength = 100,
  quantile_j2,
  quantile_rest = 0.5,
  ...
)

Arguments

NLmod	The fitted model
X	an n by p matrix of exposures to evaluate. These should be continuous.
C	An n by q matrix of additional covariates to adjust for.
j1	The exposure for which we will plot the exposure response function
j2	The second exposure which we will fix at a certain value. This should be a secondary exposure of interest in the sense that we want to see if the exposure response curve of j1 is different at different levels of j2 (i.e interaction)
gridLength	The number of points to evaluate the curve at. More points make for a smoother curve, but also increase the computation time associated with calculating the curve
quantile_j2	The quantile at which we will fix exposure j2
quantile_rest	The quantile at which we will fix the remaining exposures
...	Plotting parameters to be passed on such as ylim, ylab, main, etc.

Examples

n = 200
p = 10
pc = 1

sigma = matrix(0.3, p, p)
diag(sigma) = 1
X = rmvnorm(n, mean=rep(0,p), sigma = sigma)

C = matrix(rnorm(n*pc), nrow=n)

TrueH = function(X) {
  return(0.5*(X[,2]*X[,3]) - 0.6*(X[,4]^2 * X[,5]))
}

Y = 5 + C + TrueH(X) + rnorm(n)

NLmod = NLint(Y=Y, X=X, C=C)

par(mfrow=c(1,2), pty='s')
plotSurface1d(NLmod = NLmod, X=X, C=C, j1=4, j2=5,
              gridLength=30, quantile_j2=0.2, quantile_rest = 0.5)
plotSurface1d(NLmod = NLmod, X=X, C=C, j1=4, j2=5,
              gridLength=30, quantile_j2=0.8, quantile_rest = 0.5)

jantonelli111/NLinteraction documentation built on Aug. 7, 2020, 10:02 p.m.