cbrown5

#Apply scale to just numeric variables

#keep all variables
dat2 <- dat %>%
    mutate(across(where(is.numeric), scale))

#keep just numerical variables, drop the others from the data.frame
dat2 <- dat %>%
    transmute(across(where(is.numeric), scale))

cbrown5

#Demonstration of splines in R
#Can be used to add nonlinearity to linear models (e.g. GLMs)

library(splines)
x <- rnorm(10000)
n_df <- 5
ns_design_mat <- ns(x, n_df) #n_df degrees of freedom
plot(x, ns_design_mat[,1], ylim = c(-1, 1))
points(x, ns_design_mat[,2], col = "red")
points(x, ns_design_mat[,3], col = "blue")

cbrown5

# Template for doing empirical Bayes prediction from a GAM with mgcv
# CJ Brown 2021-08-23
#
# See Wood 2017 and ?predict.gam for more info

#First of all its often easiest to create your prediction dataframe with
# visreg, e.g. this makes a df across all values of NEOLI_Total
# and "has_fee": 
xfit <-  visreg(tour_mod, xvar = "NEOLI_Total", partial = FALSE, 
                plot = FALSE, by = "has_fee")

cbrown5

#Functions to simulate predictions and CIs from a GAM
# CJ Brown 
# 2023-03-24
# 
# based on code in ?gam.predict and  Wood, S.N. (2017) Generalized Additive Models: An
# Introduction with R (2nd edition). Chapman and
# Hall/CRC.

library(lazyeval)

cbrown5

#Simulate from an RW2 (random walk order 2) model.
# Useful for generating predictions from INLA RW2 models. 
# When forecast, RW2 models predict continuation of trend, with 
# deviations from the trend controlled by the sd parameter.
#
# See: https://inla.r-inla-download.org/r-inla.org/doc/latent/rw2.pdf
# Can intialize the RW2 at a given starting point, this needs
# needs to be two consecutive points, given we are working with
# second order differences.