Std_report_weekStart_16March.Rmd

---
title: "ccc"
author: "Pierre Nouvellet"
date: "2019-07"
---

```{r options, include = FALSE, message = FALSE, warning = FALSE, error = FALSE}
library(knitr)
library(Hmisc)

opts_chunk$set(collapse = TRUE)

opts_chunk$set(fig.path='figs/', fig.keep='high', 
               dev=c('png'), fig.width=8, fig.height=8, cache=FALSE,
               tidy=FALSE, warning=FALSE, fig.show="asis"
 	       )

```

# models included in output

```{r}
# models
models <- list(week1 = c('RtI1','RtI0'),
               week2 = c('RtI1','RtI0'))
n_models <- lapply(models, function(x) length(x))

# time of end of weeks before predictions
day_end_week <- as.Date(c('2020-03-08','2020-03-15'),format = '%Y-%m-%d')
n_pred <- length(day_end_week)

if (n_pred != length(models)) print('warning')

```

# Open various prediction files

```{r}
preds <- list()
for (j in 1:n_pred){
   preds[[as.character(day_end_week[j])]] <- list()
  for (i in 1:n_models[[j]]){
    preds[[j]][[models[[j]][i]]] <- readRDS(
      paste0('RData/',models[[j]][i],'_Std_results_week_end_',day_end_week[j],'.rds'))
  }
}



```

# Combine results

### how many unique countries/models/forecasted days included overall

```{r}

# number of unique models across all predictions
ind_models <- sort(unique(unlist(models)))

# number of prediction days = 7*n_pred

# total number of country acroos all predictions
ind_country <- preds[[1]][[1]]$Country
for (i in 2:n_pred){
  ind_country <- c(ind_country,preds[[i]][[1]]$Country)
}
ind_country <- sort(unique(ind_country))

```

### combine Rts

```{r}

Rts <- list()
for (i in 1:n_pred){
  temp <- as.data.frame(matrix(NA,length(ind_models)*1e4,length(ind_country)+1))
  names(temp) <- c('model',ind_country)
  temp$model <- rep(ind_models,each = 1e4)
  
  f_models <- 
  
  Rts[[as.character(day_end_week[i])]] <-
}

```






## below is old stuff, i was trying, check in case useful 
a lot of it is organising the ouput of the different models/predictions week in a format easy to handle...

```{r}
# extract and combine Rts last
R_last <- cbind(model = rep(models[[1]][1],1e4),preds[[n_pred]][[1]]$Rt_last)
for (i in 2:n_models[[n_pred]]){
  R_last <- rbind(R_last,preds[[n_pred]][[i]]$Rt_last)
}
R_last_summary <- apply(R_last,2,quantile,c(.5,.025,.975))
country <- preds[[n_pred]][[1]]$Country
# clean country names
country[which(country == 'Saudi_Arabia')] <- 'S.Arabia'
country[which(country == 'United_States_of_America')] <- 'USA'
country[which(country == 'United_Kingdom')] <- 'UK'
country[which(country == 'Czech_Republic')] <- 'Czech R.'
country[which(country == 'South_Korea')] <- 'S.Korea'
country[which(country == 'San_Marino')] <- 'S.Marino'

N_geo <- length(country)

# plot errorbar of Rt last
## margin for side 2 is 7 lines in size
op <- par(mar = c(6,4,4,2) + 0.1) ## default is c(5,4,4,2) + 0.1

errbar(1:N_geo,R_last_summary[1,],R_last_summary[2,],R_last_summary[3,],
       xlab = '', ylab = 'R',ylim = c(0,8), bty = 'n',xaxt = "n")
lines(c(1,N_geo),rep(1,2), col = 'red')

axis(1, at=1:N_geo, labels=country,las=2)

# extract and combine predictions
Prediction <- list()
Prediction <- preds[[1]][[1]]$Predictions

for (i in 2:n_models[[n_pred]]){
  for (c in 1:length(preds[[1]][[1]]$Predictions)){
    f <- which(names(preds[[1]][[1]]$Predictions)[c] %in% names(Prediction))
    
  }
    Prediction[[c]] <- rbind(Prediction[[c]]$Predictions,preds[[1]][[1]]$Predictions)
}
R_last_summary <- apply(R_last,2,quantile,c(.5,.025,.975))

# get daily and weekly median and 95%CrI for all prediction weeks

# save summary incl. for each predictions weeks per country:
### Rt (median, 95%CrI); 
### observed weekly count (if available); 
### predicted weekly count; 
### performance for that week



# plot for each country the incidence oberved and the various predictions, 
### grey for past predictions, blue for news



```


# Indvidual model results

```{r}
# extract Rts last


# plot errorbar of Rt last, 1 color per model

# extract predictions

# get daily and weekly median and 95%CrI for all prediction weeks per model

# save summary incl. for each retrospective predictions weeks per country:
### performance for each week retrospective



# plot for each country the incidence oberved and the various predictions, 
### 1 color per model


```