Skip to contents

Exploring DLNM sensitivity in `cityClimateHealth`

dlnm_sensitivity.Rmd

This is where you can do the dlnm_sweep stuff

DLNM Sweep

DLNM Sweep uses a custom function called make_cb_list() to expand crossbasis options:

# cb2 <- make_cb_list(x = lndn,
#                     var_list = c('tmean'), 
#                     argvar_list = list(set1 = list(fun="bs", degree = 2, knots = list(vk))) ,
#                     lag_list = maxlag,
#                     arglag_list = list(set1 = list(knots = list(lk))))

You can see that this creates the same cb object as in the basic DLNM (with some novel attributes removed):

# attr(cb2[[1]], 'summary_str') <- NULL
# attr(cb2[[1]], 'var_name') <- NULL
# identical(cb, cb2[[1]])

Ok so now how to use this to its fullest potential. One application might look like this:

# var_list <- c('tmean', 'tmin')
# 
# cb_all <- list()
# 
# for(var in var_list) {
#   
#   # make var knots that vary by var
#   vk2 = equalknots(lndn[, var], fun="bs", degree = 2, df = 4)
#   vk3 = equalknots(lndn[, var], fun="ns", df = 4)
#   k50 = quantile(lndn[, var], probs = .5, na.rm = T)
#   k10 = quantile(lndn[, var], probs = c(1:9)/10, na.rm = T)
#   
#   cb_var = make_cb_list(x = lndn,
#                          var_list = var, 
#                          argvar_list = list(set1 = list(fun="bs", degree = c(2, 3), 
#                                                         knots = list(vk2, k50, k10)),
#                                             set2 = list(fun="ns",  
#                                                         knots = list(vk3, k50, k10))) ,
#                          lag_list = maxlag,
#                          arglag_list = list(set1 = list(knots = list(lk))))
#   
#   cb_all <- append(cb_all, cb_var)
# }