| refcurv {refreg} | R Documentation |
Univariate reference curve model
Description
This function obtain univariate conditional quantiles as described in Martinez-Silva et. al (2016).
Usage
refcurv(mu = "y~s(x)", sigma = "~s(x)", data = data)
Arguments
mu |
A formula object for the response mean model following the mgcv package structure (see example below). |
sigma |
a formula object for fitting a model to the response variance (see example below). |
data |
A data frame containing both the response, and predictor variables. |
Details
In the Martinez Silva et. al (2016) the non linear effects of the continuous covariates are estimating through polynomial kernel smoother, in this package we implement the same methodology but using penalized splines in order to reduce computational cost.
Value
This function returns univariate conditional quantiles estimated using a non parametric location scale model.
References
Martinez–Silva, I., Roca–Pardinas, J., & Ordonez, C. (2016). Forecasting SO2 pollution incidents by means of quantile curves based on additive models. Environmetrics, 27(3), 147–157.
Examples
#--- Glycation hemoglobin reference curve depending on age
dm_no <- subset(aegis, aegis$dm == "no")
fit1 <- refcurv(mu = "hba1c~s(age)", sigma = "~s(age)", data = dm_no)
plot(fit1, newdata = data.frame(age = 18:90), tau = c(0.025, 0.05, 0.10, 0.90, 0.95, 0.975))