From 747e7665713080adc3bb8c331b79ea61a2a38666 Mon Sep 17 00:00:00 2001
From: chitrams <13806861+chitrams@users.noreply.github.com>
Date: Mon, 25 Nov 2024 11:58:47 +0800
Subject: [PATCH 1/2] Started discussion of other surveys in vignette, #151

---
 vignettes/other-data-sources.Rmd | 16 +++++++++++++---
 1 file changed, 13 insertions(+), 3 deletions(-)

diff --git a/vignettes/other-data-sources.Rmd b/vignettes/other-data-sources.Rmd
index 51b92503..62996ec8 100644
--- a/vignettes/other-data-sources.Rmd
+++ b/vignettes/other-data-sources.Rmd
@@ -20,11 +20,13 @@ library(conmat)
 
 The primary goal of the conmat package is to be able to get a contact matrix for a given age population. It was initially written for work done in Australia, and so the initial focus was on cleaning and extracting data from the Australian Bureau of Statistics. 
 
-This vignette focusses on using other data sources with conmat.
+This vignette focuses on using other data sources with conmat.
+
+# Using `socialmixr`
 
 We can use some other functions from `socialmixr` to extract similar estimates for different populations in different countries.
 
-We could extract some data from Italy using the [`socialmixr`](https://epiforecasts.io/socialmixr/) R package
+We could extract some data from Italy using the [`socialmixr`](https://epiforecasts.io/socialmixr/) R package:
 
 ```{r}
 library(socialmixr)
@@ -34,7 +36,7 @@ italy_2005 <- wpp_age("Italy", "2005")
 head(italy_2005)
 ```
 
-We can then convert this data into a `conmat_population` object and use it in `extrapolate_polymod`
+We can then convert this data into a `conmat_population` object and use it in `extrapolate_polymod`:
 
 ```{r}
 italy_2005_pop <- as_conmat_population(
@@ -54,6 +56,14 @@ italy_contact <- extrapolate_polymod(
 italy_contact
 ```
 
+# Fitting to other population demographics
+
+It is important to consider the contact survey you are using to fit to your country or population of interest.
+
+Models built on contact patterns (and thus contact surveys) different to your population of interest may give different results. For example, if a model was fitted using the POLYMOD contact survey (which is based in Europe, and thus generally does not have people from multiple generations living within the same household) for a population that has multi-generational households such as China, the results will likely be different in comparison to using a contact survey from China itself.
+
+For further discussion on this problem, refer to the paper "Apparent structural changes in contact patterns during COVID-19 were driven by survey design and long-term demographic trends" by Harris et al. (2024). https://arxiv.org/abs/2406.01639
+
 # Creating a next generation matrix (NGM)
 
 To create a next generation matrix, you can use either a conmat population

From 8fb1f932e1182d9ca6973364aa7477d5f74a1f8f Mon Sep 17 00:00:00 2001
From: chitrams <13806861+chitrams@users.noreply.github.com>
Date: Tue, 26 Nov 2024 09:12:58 +0800
Subject: [PATCH 2/2] Start to add China example to other data sources, #151

---
 vignettes/other-data-sources.Rmd | 37 ++++++++++++++++++++++++++++++--
 1 file changed, 35 insertions(+), 2 deletions(-)

diff --git a/vignettes/other-data-sources.Rmd b/vignettes/other-data-sources.Rmd
index 62996ec8..17d063d6 100644
--- a/vignettes/other-data-sources.Rmd
+++ b/vignettes/other-data-sources.Rmd
@@ -24,14 +24,14 @@ This vignette focuses on using other data sources with conmat.
 
 # Using `socialmixr`
 
-We can use some other functions from `socialmixr` to extract similar estimates for different populations in different countries.
+We can use some functions from `socialmixr` to extract similar estimates for different populations in different countries.
 
 We could extract some data from Italy using the [`socialmixr`](https://epiforecasts.io/socialmixr/) R package:
 
 ```{r}
 library(socialmixr)
 
-italy_2005 <- wpp_age("Italy", "2005")
+italy_2005 <- socialmixr::wpp_age("Italy", "2005")
 
 head(italy_2005)
 ```
@@ -64,6 +64,39 @@ Models built on contact patterns (and thus contact surveys) different to your po
 
 For further discussion on this problem, refer to the paper "Apparent structural changes in contact patterns during COVID-19 were driven by survey design and long-term demographic trends" by Harris et al. (2024). https://arxiv.org/abs/2406.01639
 
+We will walk through model creation using other contact surveys in this example. Here we use China. 
+
+```{r}
+# Another way of downloading the contact survey from socialmixr
+socialmixr::list_surveys()
+
+# Once we know which survey we want, we download it from Zenodo
+china_survey <- socialmixr::get_survey("https://doi.org/10.5281/zenodo.3878754")
+
+china_imputed <- impupte_contact_data(china_survey)
+
+china_filtered <- china_imputed %>%
+    dplyr::group_by(part_id) %>%
+    dplyr::mutate(
+      missing_any_contact_age = any(is.na(cnt_age_exact)),
+      missing_any_contact_setting = any(
+        is.na(cnt_home) | 
+          is.na(cnt_work) | 
+          is.na(cnt_school) | 
+          is.na(cnt_transport) | 
+          is.na(cnt_otherplace) | 
+          is.na(cnt_otherpublicplace)
+      )
+    ) %>%
+    dplyr::ungroup() %>%
+    dplyr::filter(
+      !is.na(part_age),
+      !missing_any_contact_age,
+      !missing_any_contact_setting
+    )
+```
+
+
 # Creating a next generation matrix (NGM)
 
 To create a next generation matrix, you can use either a conmat population