9 Appendix B: The Holiday Scenario

In this chapter we want to further scrutinize the dependence of each estimator on the chosen radio network specification. In practice this means we will generate a second ground truth population, called GTP2, which will mimic a scenario where people travel to hotspots, for example their work, their holiday or an event. This means that there will be an unusual high number of mobile phones in certain places that might not have the respective antenna infrastructure one would expect with such a high density of phones. Remember, the original radio network specification is largely planned on daily life activities, meaning that urban areas have a higher density of antennas and rural areas a lower one. However, if it is the holiday season more people might be traveling to rural places and therefore the antennas’ registered phones in these places might be unusual high. We want to see which estimator performs best in these situations.

For this we need to create a GTP2 that largely has a different geographical distribution than the the first ground truth population (GTP1). The radio network will stay the same however, there will be another device to cell association leading to the fact that we will end up with a new specified c.vector. We will then run all 4 estimators again with the new c.vector, the same P.matrix specification (if necessary) and compare the estimations to GTP2 with the same evaluation techniques from before.

For this chapter we need the following packages:

library(data.table) 
library(tidyverse)
library(sf)
library(raster)
library(furrr)
library(stars)
library(osc)
library(knitr)
library(DT)
library(ggthemes)
library(ggsn)
set.seed(762)

# census.classified.final.sf <- readRDS("C:/Users/Marco/Vysoká škola ekonomická v Praze/Tony Wei Tse Hung - YAY/working objects/census.classified.final.sf.rds")