inst/shiny/C19/server/3module/analysis.R
In marzione00/COVID_19_HACK: Covid_19 outbreak analysis
## REACTIVES OF THIS CODE
reac_general <- shiny::reactiveValues()
reac_logistic <- shiny::reactiveValues()
reac_ARIMA <- shiny::reactiveValues(arimaOK = FALSE)
reac_FFT <- shiny::reactiveValues()
reac_R <- shiny::reactiveValues()
reac_SEIR <- shiny::reactiveValues()
# THIS IS THE REACTIVE CONTAINER OF TERRITORY SELECTION VARIABLES.
t <- shiny::reactiveValues()
## Reactive values for global storage
# Use this function waitLoading() to validate the content of t$r and prevent errors
# (just use it at the beginning of each chunk of code)
waitLoading <- shiny::reactive({
shiny::validate(
shiny::need(is_ready(t$r), "Wait...")
)
return(t$r)
})
waitInput <- shiny::reactive({
shiny::validate(
shiny::need(is_ready(input$region), "Wait...")
)
return(input$region)
})
# DEPENDS ON REACTIVE t
suggest_dates <- function()
{
index = covid19::const_trim(eval(t$data)[[t$name]]$totale_casi,1)
sugStart = eval(t$data)[[t$name]]$data[index]
#sugStart = reac_general$sample_date[index]
return(c(sugStart, fin_date))
}
# DEPENDS ON REACTIVES t AND reac_ARIMA! (but not on reac_ARIMA$arima)
suggest_lags <- function()
{
checkExp({
sugg <- toString(forecast::auto.arima(log(reac_ARIMA$sample_cases_trim)))
matches <- regmatches(sugg, gregexpr("[[:digit:]]+", sugg))
return(as.numeric(unlist(matches)))
}, "There is no suitable ARIMA model")
}
# Province input updater
shiny::observe({
wait <- waitInput()
if(input$region != "") {
ch <- c("--- ALL ---" = "", regAndProv[regAndProv$region == input$region, "province"])
} else {
ch <- c("--- ALL ---" = "", provNames)
}
shiny::updateSelectizeInput(session, inputId = "province", choices = ch, selected = NULL)
})
# Assignment of territory variables when pressing action button
shiny::observe( {
if(input$region == "" && input$province == "") {
t$name <- input$country
t$data <- expression(countryTS)
t$c <- TRUE
t$r <- FALSE
t$p <- FALSE
} else if(input$province != "") {
t$name <- input$province
t$data <- expression(provTS)
t$c <- FALSE
t$r <- FALSE
t$p <- TRUE
} else {
t$name <- input$region
t$data <- expression(regionTS)
t$c <- FALSE
t$r <- TRUE
t$p <- FALSE
}
reac_general$sugg_dates = suggest_dates()
shiny::updateSliderInput(session,"arima_interval",min = init_date, max = fin_date, timeFormat = "%d %b",
step = 1, value = c(reac_general$sugg_dates[1], reac_general$sugg_dates[2]))
shiny::updateSliderInput(session, "fitInterval",min = init_date, max = fin_date, timeFormat = "%d %b",
step = 1, value = c(reac_general$sugg_dates[1],reac_general$sugg_dates[2]))
shiny::updateSliderInput(session,"FFT_interval",min = init_date, max = fin_date, timeFormat = "%d %b",
step = 1, value = c(reac_general$sugg_dates[1], reac_general$sugg_dates[2]))
})
## REGION plot (currently date against total cases) ##
output$dates_sugg <- shiny::renderUI({
wait <- waitLoading()
if( is_ready(reac_general$sugg_dates[1]) ) {
h4(paste("Suggested initial and final date: ", reac_general$sugg_dates[1]," - ", reac_general$sugg_dates[2]))
}
})
shiny::observe({
wait <- waitLoading()
reac_general$sample_date <- eval(t$data)[[t$name]]$data
reac_general$sample_cases <- covid19:::level_and_fill(eval(t$data)[[t$name]]$totale_casi, direction = input$direction_fill, method = input$method_fill)
})
output$coolplot1 <- plotly::renderPlotly({
wait <- waitLoading()
#waiter::waiter_show(id = "coolplot1", html = waiter::spin_loaders(id = 1, color = "#ff471a"), color = "white")
# Data trim and curve fitting #
if(is_ready(reac_general$sample_cases)) {
n <- nrow(eval(t$data)[[t$name]])
logic_interval <- reac_general$sample_date >= input$fitInterval[1] &
reac_general$sample_date <= input$fitInterval[2]
sample_serial_date <- eval(t$data)[[t$name]]$data_seriale
sample_cases <- reac_general$sample_cases
sample_diff <- c(NA,diff(sample_cases))
if( !t$p && input$swab_std ) {
swabs <- covid19:::level_and_fill(eval(t$data)[[t$name]]$tamponi, direction = input$direction_fill, method = input$method_fill)
sample_cases <- sample_cases / swabs
sample_diff <- c(NA,sample_diff[-1] / diff(swabs))
}
sample_serial_date_trim <- sample_serial_date[logic_interval]
sample_cases_trim <- sample_cases[logic_interval]
sample_diff_trim <- sample_diff[logic_interval]
sample_serial_date_rem <- sample_serial_date[!logic_interval]
sample_cases_rem <- sample_cases[!logic_interval]
sample_diff_rem <- sample_diff[!logic_interval]
fit_data <- covid19:::exe_fit(sample_cases = sample_cases_trim,
sample_date = sample_serial_date_trim,
days = days)
reac_logistic$model <- fit_data$out_fit$model
reac_logistic$resid <- fit_data$out_resid
reac_logistic$vals <- fit_data$out_fit$vals
conf <- nlstools::confint2(level = 0.95, object = reac_logistic$model)
yConf_up <- (conf["n0",2]*conf["k",2])/(conf["n0",2] + (conf["k",2]-conf["n0",2]) * exp(-conf["r",2]*(sample_serial_date_trim-sample_serial_date_trim[1]))) + sample_cases_trim[1]
yConf_down <- (conf["n0",1]*conf["k",1])/(conf["n0",1] + (conf["k",1]-conf["n0",1]) * exp(-conf["r",1]*(sample_serial_date_trim-sample_serial_date_trim[1]))) + sample_cases_trim[1]
# Conversion to real date and creation of fitted points #
points_trim <- data.frame("sample_serial_date_trim" = reac_general$sample_date[logic_interval],
sample_cases_trim)
points_rem <- data.frame("sample_serial_date_rem" = reac_general$sample_date[!logic_interval],
sample_cases_rem)
points_diff_trim <- data.frame("sample_serial_date_trim" = reac_general$sample_date[logic_interval],
sample_diff_trim)
points_diff_rem <- data.frame("sample_serial_date_rem" = reac_general$sample_date[!logic_interval],
sample_diff_rem)
fittedPoints <- fit_data$fittedPoints
fittedPoints_der <- fit_data$fittedPoints_der
seq_dates <- seq(from = init_date, by = 1, length.out = length(days))
fittedPoints$days <- seq_dates
fittedPoints_der$days <- seq_dates
confPoints_up <- data.frame("sample_serial_date_trim" = reac_general$sample_date[logic_interval],
yConf_up)
confPoints_down <- data.frame("sample_serial_date_trim" = reac_general$sample_date[logic_interval],
yConf_down)
# PLOT with plotly #
fig = plotly::plot_ly( name = "Cases", type= "scatter")
# funtions for the two different plots
plot1 = function(fig)
{
fig <- fig %>% plotly::add_trace(data = confPoints_down, x = ~sample_serial_date_trim, y = ~yConf_down, mode='none',hoverinfo='skip',fill = 'tozeroy', name="IGNORED_LEGEND", fillcolor="rgba(0,0,0,0)",showlegend = FALSE)
fig <- fig %>% plotly::add_trace(data = confPoints_up, x = ~sample_serial_date_trim, y = ~yConf_up, mode='none', fill = 'tonexty' ,name="Confidence interval 95%", fillcolor="rgb(255,250,205)")
hovlabels <- c("")
for(i in c(1:n)) {
hovlabels[i] <- paste(format(reac_general$sample_date[i], "%d %b"),
", Tot cases = ", sample_cases[i], sep ="")
if( !t$p && input$swab_std )
hovlabels[i] <- paste(hovlabels[i], ", Tot swabs = ", swabs[i], sep = "")
}
fig <- fig %>% plotly::add_trace(data = points_rem, x =~sample_serial_date_rem, y =~sample_cases_rem ,marker = list(color = "red"), mode = 'markers', name = "Total cases (excluded)",
text = hovlabels[!logic_interval], hoverinfo = 'text')
fig <- fig %>% plotly::add_trace(data = points_trim, x =~sample_serial_date_trim, y =~sample_cases_trim ,marker = list(color = "green"), mode = 'markers', name = "Total cases (fitting)",
text = hovlabels[logic_interval], hoverinfo = 'text')
fig <- fig %>% plotly::add_trace(data = fittedPoints, x = ~days, y = ~yFitted, line = list(color ='rgb(0,0,139)',width=2.5), mode='lines', name = "Fitted logistic curve" )
return(fig)
}
plot2 = function (fig)
{
hovlabels <- c("")
for(i in c(2:n)) {
hovlabels[i] <- paste(format(reac_general$sample_date[i], "%d %b"),
", Cases = ",
sample_cases[i] - sample_cases[i-1],
sep ="")
if( !t$p && input$swab_std )
hovlabels[i] <- paste(hovlabels[i], ", Swabs = ", swabs[i]-swabs[i-1], sep = "")
}
fig <- fig %>% plotly::add_bars(data = points_diff_rem, x =~sample_serial_date_rem, y =~sample_diff_rem, marker = list(color = "red"), name = "New cases (excluded)",
text = hovlabels[!logic_interval], hoverinfo = 'text')
fig <- fig %>% plotly::add_bars(data = points_diff_trim, x =~sample_serial_date_trim, y =~sample_diff_trim, marker = list(color = "green"), name = "New cases (fitting)",
text = hovlabels[logic_interval], hoverinfo = 'text')
fig <- fig %>% plotly::add_trace(data = fittedPoints_der, x = ~days, y = ~yFitted_der, line = list(color ='rgb(255,117,20)',width=2.5), mode='lines', name= "Fitted logistic distribution")
return(fig)
}
#Plot based on the checkbox
if( 1 %in% input$plot_type && !(2 %in% input$plot_type ) )
{
fig = plot1(fig)
} else if( 2 %in% input$plot_type && !(1 %in% input$plot_type ) )
{
fig = plot2(fig)
} else if( 1 %in% input$plot_type && (2 %in% input$plot_type ) )
{
fig = plot1(fig)
fig = plot2(fig)
}
# labels and plot
fig <- fig %>%plotly::layout(xaxis = list(title = "day"), yaxis = list(title = "Infected"))
if( reac_logistic$vals$k > 1e7 )
fig <- fig %>%plotly::layout(title = "Warning: unrealistic model estimated", font = list(color = 'red'),dtick= 5 )
fig
}
})
#-- Summary of regions ---
output$fit_smry <- shiny::renderPrint({
wait <- waitLoading()
summary(reac_logistic$model)
})
output$selected_terr <- shiny::renderPrint({
})
selected_info_func <- function() {
if(is_ready(t$name))
{
div(h2(strong(paste("Selected territory:", t$name))), align = "center", style = "color:red")
}
}
output$selected_info1 <- shiny::renderUI({
selected_info_func()
})
output$selected_info2 <- shiny::renderUI({
selected_info_func()
})
output$selected_info3 <- shiny::renderUI({
selected_info_func()
})
output$selected_info4 <- shiny::renderUI({
selected_info_func()
})
output$selected_info5 <- shiny::renderUI({
if(is_ready(reac_SEIR$name) && is_ready(reac_SEIR$notAvail))
{
if(reac_SEIR$notAvail) {
fluidPage(
div(h2(strong(paste("Selected territory:", reac_SEIR$name))), align = "center", style = "color:red"),
div(h5(em(paste("Data for SEIR not available at province level"))), align = "center", style = "color:red")
)
} else {
div(h2(strong(paste("Selected territory:", reac_SEIR$name))), align = "center", style = "color:red")
}
}
})
output$resid_smry <- shiny::renderPrint({
wait <- waitLoading()
print("Data: residuals from nls regression")
nlstools::test.nlsResiduals(reac_logistic$resid)
})
#-- Residuals ---
output$plot_residual <- plotly::renderPlotly({
wait <- waitLoading()
Res_DF_1<-as.data.frame(reac_logistic$resid$resi1)
Res_DF_2<-as.data.frame(reac_logistic$resid$resi2)
Res_DF_3<-as.data.frame(reac_logistic$resid$resi4)
Res_DF_4<-as.data.frame(reac_logistic$resid$resi3)
p = plotly::plot_ly(type = 'scatter')
p <- p %>% plotly::layout(
xaxis = list(
showline = FALSE,
dtick = 2000,
zeroline = FALSE
),
yaxis = list(
showline = FALSE,
zeroline = FALSE
)
)
if( is_ready(input$plot_res_type) ) {
if(input$plot_res_type=="Residuals"){
colnames(Res_DF_1)=c("fitted1","res")
p <- p %>% plotly::layout(
title ="Residuals",
xaxis = list(title="Fitted values",zeroline = FALSE),
yaxis = list(title="Residuals")
)
p= p %>%plotly::add_trace(name = "residual",showlegend=FALSE,data=Res_DF_1,x=~Res_DF_1$fitted1,y=~Res_DF_1$res,marker = list(size = 15,
color = 'rgba(255, 182, 193, .9)',
line = list(color = 'rgba(152, 0, 0, .8)',
width = 2)))
p <- p %>%plotly::add_trace(data=Res_DF_2,x=~Res_DF_1$fitted1,y = 0, name = "liney0", line = list(color = 'rgb(0,0,0)', width = 1, dash = 'dot'),showlegend = FALSE, mode = 'lines')
p
#grafico1
}
else if(input$plot_res_type=="Residuals_standardized"){
colnames(Res_DF_2)=c("fitted2","res_stand")
p <- p %>% plotly::layout(
title ="Residuals standardized",
xaxis = list(title="Fitted values"),
yaxis = list(title="Standardized residuals")
)
p <- p %>%plotly::add_trace(name="Standardised residuals",showlegend=FALSE,data=Res_DF_2,x=~Res_DF_2$fitted2,y=~Res_DF_2$res_stand,marker = list(size = 15,
color = 'rgba(255, 182, 193, .9)',
line = list(color = 'rgba(152, 0, 0, .8)',
width = 2)))
p <- p %>%plotly::add_trace(data=Res_DF_2,x=~Res_DF_2$fitted2,y = 0, name = "liney0", line = list(color = 'rgb(0,0,0)', width = 1, dash = 'dot'),showlegend = FALSE, mode = 'lines')
p <- p %>%plotly::add_trace(data=Res_DF_2,x=~Res_DF_2$fitted2,y = 2, name = 'liney2', line = list(color = 'rgb(0,0,0)', width = 1),showlegend = FALSE ,mode = 'lines')
p
}
else if(input$plot_res_type=="Autocorrelation"){
colnames(Res_DF_3)=c("fitted3","resiplus1")
p <- p %>% plotly::layout(
xaxis = list(title="Residuals i"),
yaxis = list(title="Residuals i+1"),
title ="Autocorrelation")
p = p %>%plotly::add_trace(name = "Autocorrelation", showlegend=FALSE,data=Res_DF_3,x=~Res_DF_3$fitted3,y=~Res_DF_3$resiplus1,marker = list(size = 15,
color = 'rgba(255, 182, 193, .9)',
line = list(color = 'rgba(152, 0, 0, .8)',
width = 2)))
p <- p %>%plotly::add_trace(data=Res_DF_3,x=~Res_DF_3$fitted3,y = 0, name = "liney0", line = list(color = 'rgb(0,0,0)', width = 1, dash = 'dot'),showlegend = FALSE, mode="lines")
p
#grafico1
}
else if(input$plot_res_type=="Sqrt of abs of res vs fitted"){
p <- p %>% plotly::layout(
title ="Sqrt of abs of residual",
xaxis = list(title="Fitted"),
yaxis = list(title="Residuals")
)
colnames(Res_DF_4)=c("fitted4","qq")
p = p %>%plotly::add_trace(name="Sqrt of abs of res vs fitted",showlegend=FALSE,data=Res_DF_4,x=~Res_DF_4$fitted4,y=~Res_DF_4$qq,marker = list(size = 15,
color = 'rgba(255, 182, 193, .9)',
line = list(color = 'rgba(152, 0, 0, .8)',
width = 2)))
p
#grafico1
}
}
})
##===================================== ARIMA SECTION =============================================##
## HERE GOES ALL ARIMA IMPLEMENTATION COMMON TO ALL GRAPHS
shiny::observe({
wait <- waitLoading()
if(is_ready(input$arima_interval[1]) && is_ready(reac_general$sample_cases)) {
reac_ARIMA$logic_interval <- reac_general$sample_date >= input$arima_interval[1] &
reac_general$sample_date <= input$arima_interval[2]
reac_ARIMA$sample_date_trim <- reac_general$sample_date[reac_ARIMA$logic_interval]
reac_ARIMA$sample_cases_trim <- reac_general$sample_cases[reac_ARIMA$logic_interval]+1
reac_ARIMA$sugg_lags <- suggest_lags()
updateSliderInput(session,"ARIMA_p", value= reac_ARIMA$sugg_lags[1])
updateSliderInput(session,"ARIMA_I", value = reac_ARIMA$sugg_lags[2])
updateSliderInput(session,"ARIMA_q", value = reac_ARIMA$sugg_lags[3])
}
})
shiny::observe({
wait <- waitLoading()
if(is_ready(input$ARIMA_p)) {
reac_ARIMA$arima <- checkExp( stats::arima(log(reac_ARIMA$sample_cases_trim),order=c(input$ARIMA_p,input$ARIMA_I,input$ARIMA_q)) , "There is not a suitable ARIMA model")
reac_ARIMA$arimaOK <- TRUE
}
})
#Ln x = Log10 x / Log10 e
## Plot of ARIMA Time Series and FORECAST
output$arima_coolplot1 <- plotly::renderPlotly({
wait <- waitLoading()
#acf(log(sample_cases_trim))
#pacf(log(sample_cases_trim))
if( reac_ARIMA$arimaOK ) {
fore = checkExp(forecast::forecast(reac_ARIMA$arima,input$forecast) , "There is not a forecast for the ARIMA model")
sdt <- reac_ARIMA$sample_date_trim
fore.dates <- seq(from = sdt[length(sdt)], by = 1, len = input$forecast)
p <-plotly::plot_ly() %>%
plotly::add_ribbons(x = fore.dates,
ymin = exp(fore$mean),
ymax = exp(fore$upper[, 2]),
color = I("#17becf"),
name = "95% confidence") %>%
plotly::add_ribbons(p,
x = fore.dates,
ymin = exp(fore$mean),
ymax = exp(fore$upper[, 1]),
color = I("#ed9dac"), name = "80% confidence")%>%
plotly::add_lines(x = sdt, y = reac_ARIMA$sample_cases_trim,
color = I("#037d50"),
name = "observed",
mode="lines")%>%
plotly::add_lines(x = fore.dates,
y = exp(fore$mean),
color = I("#ee1147"),
name = "prediction")
p <- p %>% plotly::layout(
title = paste0("ARIMA Forecast (",input$ARIMA_p,",",input$ARIMA_I,",",input$ARIMA_q,")"),
xaxis = list(title="Days"),
yaxis = list(title="Total cases", type = "log")
)
p
}
})
## Plot of autocorrelation function
output$arima_coolplot2 <- plotly::renderPlotly({
wait <- waitLoading()
if(is_ready(reac_ARIMA$sample_cases_trim)) {
p = checkExp( forecast::autoplot(acf(log(reac_ARIMA$sample_cases_trim))) , "There is not a forecast for the ARIMA model")
p = plotly::ggplotly(p)
p <- p %>%plotly::layout(xaxis = list(title = "LAG"), yaxis = list(title = "ACF"), title="Autocorrelation", showlegend = TRUE,
plot_bgcolor = "rgb(255, 255, 255)")
}
})
## Plot of partial autocorrelation function
output$arima_coolplot3 <- plotly::renderPlotly({
wait <- waitLoading()
if(is_ready(reac_ARIMA$sample_cases_trim)) {
p = checkExp( forecast::autoplot(pacf(log(reac_ARIMA$sample_cases_trim)) ) , "There is not a forecast for the ARIMA model")
p = plotly::ggplotly(p)
p <- p %>%plotly::layout(xaxis = list(title = "LAG"), yaxis = list(title = "PACF"), title="Partial Autocorrelation", showlegend = TRUE,
plot_bgcolor = "rgb(255, 255, 255)")
}
})
## Plot of ARIMA residuals
output$arima_coolplot4 <- shiny::renderPlot({
wait <- waitLoading()
if(reac_ARIMA$arimaOK) {
checkExp(forecast::checkresiduals(reac_ARIMA$arima) , "There is not a suitable ARIMA model")
}
})
# --- Summary ARIMA
## Print of suggested parameters
output$parameters_sugg <- shiny::renderUI({
wait <- waitLoading()
if( is_ready(reac_ARIMA$sugg_lags[1]) ) {
fluidPage(
fluidRow(
h3( paste("Suggested Parameters: ARIMA(", paste(reac_ARIMA$sugg_lags, collapse = ","), ")") )
),
fluidRow(
h4(paste("Suggested initial and final date: ", reac_general$sugg_dates[1]," - ", reac_general$sugg_dates[2]))
)
)
}
})
# toString(reac_ARIMA$arima$coef)
# suggested fit toString(forecast::auto.arima(log(sample_cases_trim)))
#reac_ARIMA$arima$coef
# reac_ARIMA$arima$sigma2
output$arima_shell_output <- shiny::renderPrint({
wait <- waitLoading()
if(reac_ARIMA$arimaOK) {
reac_ARIMA$arima
}
})
output$arima_shell_resid <- shiny::renderPrint({
wait <- waitLoading()
if(reac_ARIMA$arimaOK) {
checkExp(forecast::checkresiduals(reac_ARIMA$arima,plot=FALSE), "There is not a suitable ARIMA model")
}
})
#============ FFT ======================================
shiny::observe({
wait <- waitLoading()
if(is_ready(reac_general$sample_cases)) {
reac_FFT$logic_interval <- reac_general$sample_date >= input$FFT_interval[1] &
reac_general$sample_date <= input$FFT_interval[2]
reac_FFT$sample_date_trim <- reac_general$sample_date[reac_FFT$logic_interval]
reac_FFT$sample_cases_trim <- reac_general$sample_cases[reac_FFT$logic_interval]+1
}
})
output$FFT_day_cases<- shiny::renderPlot({
if(is_ready(reac_FFT$sample_cases_trim)) {
FFTX<-spectral::spec.fft(diff(reac_FFT$sample_cases_trim))
plot(FFTX,type = "l",ylab = "Amplitude",xlab = "Frequency",lwd = 2)
}
})
output$FFT_day_cases_diff<- shiny::renderPlot({
if(is_ready(reac_FFT$sample_cases_trim)) {
FFTX<-spectral::spec.fft(diff(diff(reac_FFT$sample_cases_trim)))
plot(FFTX,type = "l",ylab = "Amplitude",xlab = "Frequency",lwd = 2)
}
})
#======================= R(T) ==================================
#DEFAULT 1.4, 0.8
#
#
#shiny::observe({
#
# if(is_ready(reac_general$sample_cases) && is_ready(input$"Gamma_1")) {
#
# GT.chld.hsld2<-R0::generation.time("gamma", c(input$"Gamma_1", input$"Gamma_2"))
# time_R <- as.numeric(length(reac_general$sample_cases))-2
#
# tryCatch(
# {
# reac_R$R0_data <- R0::est.R0.TD(diff(reac_general$sample_cases),GT.chld.hsld2, begin=1, end=time_R)
# },
# error = function(e)
# {
# reac_R$R0_data <- FALSE
# }
# )
# }
#})
#
#R0_data <- shiny::reactive({
#
# shiny::validate(
# shiny::need(is_ready(reac_R$R0_data), "Wait...") %then%
# shiny::need(!isFALSE(reac_R$R0_data), "The estimates cannot be calculated")
# )
# return(reac_R$R0_data)
#})
#
#
#
#output$R_t_evaluation<- shiny::renderPlot({
#
# plot( R0_data())
#
#})
#
#output$R_t_goodness_of_fit<- shiny::renderPlot({
#
# R0::plotfit(R0_data())
#
#})
#
#output$R_t_evaluation_FFT<- shiny::renderPlot({
#
# R0_data_raw_FFT <- spectral::spec.fft(R0_data()[["R"]])
# plot(R0_data_raw_FFT,type = "l",ylab = "Amplitude",xlab = "Frequency",lwd = 2)
#
#})
#======================= SEIR MODEL ==================================
#
#
# #---- data retrieval, common to both
#
# distr_IT <- 'exp'
# distr_SRT <- 'exp'
# distr_IBST <- 'none'
#
# shiny::observe({
#
# wait <- waitLoading()
#
# # territory selection for SEIR
# if(t$p) {
# reac_SEIR$data <- expression(regionTS)
# reac_SEIR$notAvail <- TRUE
# reac_SEIR$name <- regAndProv[ regAndProv$province == t$name, "region"]
# } else {
# reac_SEIR$data <- t$data
# reac_SEIR$notAvail <- FALSE
# reac_SEIR$name <- t$name
# }
#
# # N = population of area
# if(t$c) {
# reac_SEIR$N <- italy_pop$country[1,"valore"]
# } else {
# reac_SEIR$N <- italy_pop$region[italy_pop$region$territorio == reac_SEIR$name, "valore"]
# }
#
# ## Time series
# # Infected Time Series
# reac_SEIR$P <- level_and_fill(eval(reac_SEIR$data)[[reac_SEIR$name]]$totale_casi, direction = input$direction_fill, method = input$method_fill)
#
# # Removed Time Series (recovered + dead)
# reac_SEIR$R <- level_and_fill(eval(reac_SEIR$data)[[reac_SEIR$name]]$dimessi_guariti, direction = input$direction_fill, method = input$method_fill) +
# level_and_fill(eval(reac_SEIR$data)[[reac_SEIR$name]]$deceduti, direction = input$direction_fill, method = input$method_fill)
#
#
# reac_SEIR$par_IT <- list('rate'=1/input$rate_IT)
# reac_SEIR$par_SRT <- list('rate'=1/input$rate_SRT)
# reac_SEIR$par_IBST <- NULL
#
# reac_SEIR$IT <- covid19::generate(distr_IT, reac_SEIR$par_IT)
# reac_SEIR$SRT <- covid19::generate(distr_SRT, reac_SEIR$par_SRT)
# reac_SEIR$IBST <- covid19::generate(distr_IBST, reac_SEIR$par_IBST)
#
# reac_SEIR$m <- max(length(reac_SEIR$IT), length(reac_SEIR$SRT), length(reac_SEIR$IBST))
# reac_SEIR$end <- length(reac_SEIR$P) #Rt goes from day 1 to day n_obs-m BUT also the SEIR data series will go from day 1 to that day after refinement
# reac_SEIR$len <- reac_SEIR$end - reac_SEIR$m
# })
#
#
# #---- R0 part
#
# shiny::observeEvent(input$est_R0, {
#
# out <- SEIR_factotum(P = reac_SEIR$P, R = reac_SEIR$R, N = reac_SEIR$N,
# end=reac_SEIR$end, future=input$future,
# distr_IT=distr_IT, distr_SRT=distr_SRT, distr_IBST=distr_IBST,
# par_IT=reac_SEIR$par_IT, par_SRT=reac_SEIR$par_SRT, par_IBST=reac_SEIR$par_IBST,
# R0_exp_est_end=input$R0_exp_est_end,
# plot_data=input$plot_data)
#
# reac_SEIR$R0 <- out$R0
#
# S_ <- reac_SEIR$N*out$S_
# E_ <- reac_SEIR$N*out$E_
# I_ <- reac_SEIR$N*out$I_
# R_ <- reac_SEIR$N*out$R_
# S <- reac_SEIR$N*out$S
# E <- reac_SEIR$N*out$E
# I <- reac_SEIR$N*out$I
# R <- reac_SEIR$N*out$R
#
# reac_SEIR$U <- data.frame(date = seq(from = init_date, by = 1, length.out = reac_SEIR$len), 'S'=S, 'E'=E, 'I'=I, 'R'=R)
# reac_SEIR$U_ <- data.frame(date = seq(from = init_date, by = 1, length.out = reac_SEIR$len), 'S_'=S_, 'E_'=E_, 'I_'=I_, 'R_'=R_)
#
# reac_SEIR$current <- "R0"
#
# })
#
#
# shiny::observeEvent(input$est_Rt, {
#
#
# # REFERS TO INPUT IN PRECEDENT TAB
#
# GT<-R0::generation.time("gamma", c(input$"Gamma_1", input$"Gamma_2"))
# R0_out <- R0::est.R0.TD(diff(reac_SEIR$P),GT,begin=1, end=as.numeric(reac_SEIR$len))
# Rt <- R0_out$R
# R0_stages <- (1:(length(Rt)-1))
#
# out <- SEIR_factotum(P = reac_SEIR$P, R = reac_SEIR$R, N = reac_SEIR$N,
# end=reac_SEIR$end, future=input$future,
# distr_IT=distr_IT, distr_SRT=distr_SRT, distr_IBST=distr_IBST,
# par_IT=reac_SEIR$par_IT, par_SRT=reac_SEIR$par_SRT, par_IBST=reac_SEIR$par_IBST,
# R0_msp=Rt, R0_stages=R0_stages,
# plot_data=input$plot_data)
#
# S_ <- reac_SEIR$N*out$S_
# E_ <- reac_SEIR$N*out$E_
# I_ <- reac_SEIR$N*out$I_
# R_ <- reac_SEIR$N*out$R_
# S <- reac_SEIR$N*out$S
# E <- reac_SEIR$N*out$E
# I <- reac_SEIR$N*out$I
# R <- reac_SEIR$N*out$R
#
# reac_SEIR$U <- data.frame(date = seq(from = init_date, by = 1, length.out = reac_SEIR$len), 'S'=S, 'E'=E, 'I'=I, 'R'=R)
# reac_SEIR$U_ <- data.frame(date = seq(from = init_date, by = 1, length.out = reac_SEIR$len), 'S_'=S_, 'E_'=E_, 'I_'=I_, 'R_'=R_)
#
# reac_SEIR$current <- "R(t)"
#
# })
#
#
# shiny::observe({
# if(is_ready(reac_SEIR$U_) ) {
# reac_SEIR$SEIR_plot <- highcharter::hchart(tidyr::gather((reac_SEIR$U_), key="key", value="value", -date),
# highcharter::hcaes(x = date, y = value, group = key),
# type = "line",
# color = c("#bf40bf", "#e60000", "#00b300", "#fec501"),
# name = c("Exposed", "Infectious", "Recovered", "Susceptible"),
# marker = list(enabled = FALSE),
# legendIndex = c(2,3,4,1),
# tooltip = list(valueDecimals = 0)
# ) %>%
# highcharter::hc_add_series(tidyr::gather((reac_SEIR$U), key="key", value="value", -date),
# highcharter::hcaes(x = date, y = value, group = key),
# type = "scatter",
# name = c("Exposed (true)", "Infectious (true)", "Recovered (true)", "Susceptible (true)"),
# color = c("#bf40bf", "#e60000", "#00b300", "#fec501"),
# legendIndex = c(2,3,4,1),
# tooltip = list(valueDecimals = 0,
# pointFormat = '<span style="color:{point.color}">■</span> {series.name}: <b>{point.y}</b><br/>',
# headerFormat = '<span style="font-size: 10px">{point.key}</span><br/>')
# ) %>%
# highcharter::hc_title(text = paste("SEIR based on", reac_SEIR$current))
# }
# })
#
#
# output$SEIR_R0 <- renderPrint({
# if(is_ready(reac_SEIR$R0)) {
# print(reac_SEIR$R0)
# }
# })
#
# output$SEIR_plot <- highcharter::renderHighchart(
# reac_SEIR$SEIR_plot
# )
marzione00/COVID_19_HACK documentation built on Nov. 20, 2020, 1:10 p.m.
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
## REACTIVES OF THIS CODE
reac_general <- shiny::reactiveValues()
reac_logistic <- shiny::reactiveValues()
reac_ARIMA <- shiny::reactiveValues(arimaOK = FALSE)
reac_FFT <- shiny::reactiveValues()
reac_R <- shiny::reactiveValues()
reac_SEIR <- shiny::reactiveValues()
# THIS IS THE REACTIVE CONTAINER OF TERRITORY SELECTION VARIABLES.
t <- shiny::reactiveValues()
## Reactive values for global storage
# Use this function waitLoading() to validate the content of t$r and prevent errors
# (just use it at the beginning of each chunk of code)
waitLoading <- shiny::reactive({
shiny::validate(
shiny::need(is_ready(t$r), "Wait...")
)
return(t$r)
})
waitInput <- shiny::reactive({
shiny::validate(
shiny::need(is_ready(input$region), "Wait...")
)
return(input$region)
})
# DEPENDS ON REACTIVE t
suggest_dates <- function()
{
index = covid19::const_trim(eval(t$data)[[t$name]]$totale_casi,1)
sugStart = eval(t$data)[[t$name]]$data[index]
#sugStart = reac_general$sample_date[index]
return(c(sugStart, fin_date))
}
# DEPENDS ON REACTIVES t AND reac_ARIMA! (but not on reac_ARIMA$arima)
suggest_lags <- function()
{
checkExp({
sugg <- toString(forecast::auto.arima(log(reac_ARIMA$sample_cases_trim)))
matches <- regmatches(sugg, gregexpr("[[:digit:]]+", sugg))
return(as.numeric(unlist(matches)))
}, "There is no suitable ARIMA model")
}
# Province input updater
shiny::observe({
wait <- waitInput()
if(input$region != "") {
ch <- c("--- ALL ---" = "", regAndProv[regAndProv$region == input$region, "province"])
} else {
ch <- c("--- ALL ---" = "", provNames)
}
shiny::updateSelectizeInput(session, inputId = "province", choices = ch, selected = NULL)
})
# Assignment of territory variables when pressing action button
shiny::observe( {
if(input$region == "" && input$province == "") {
t$name <- input$country
t$data <- expression(countryTS)
t$c <- TRUE
t$r <- FALSE
t$p <- FALSE
} else if(input$province != "") {
t$name <- input$province
t$data <- expression(provTS)
t$c <- FALSE
t$r <- FALSE
t$p <- TRUE
} else {
t$name <- input$region
t$data <- expression(regionTS)
t$c <- FALSE
t$r <- TRUE
t$p <- FALSE
}
reac_general$sugg_dates = suggest_dates()
shiny::updateSliderInput(session,"arima_interval",min = init_date, max = fin_date, timeFormat = "%d %b",
step = 1, value = c(reac_general$sugg_dates[1], reac_general$sugg_dates[2]))
shiny::updateSliderInput(session, "fitInterval",min = init_date, max = fin_date, timeFormat = "%d %b",
step = 1, value = c(reac_general$sugg_dates[1],reac_general$sugg_dates[2]))
shiny::updateSliderInput(session,"FFT_interval",min = init_date, max = fin_date, timeFormat = "%d %b",
step = 1, value = c(reac_general$sugg_dates[1], reac_general$sugg_dates[2]))
})
## REGION plot (currently date against total cases) ##
output$dates_sugg <- shiny::renderUI({
wait <- waitLoading()
if( is_ready(reac_general$sugg_dates[1]) ) {
h4(paste("Suggested initial and final date: ", reac_general$sugg_dates[1]," - ", reac_general$sugg_dates[2]))
}
})
shiny::observe({
wait <- waitLoading()
reac_general$sample_date <- eval(t$data)[[t$name]]$data
reac_general$sample_cases <- covid19:::level_and_fill(eval(t$data)[[t$name]]$totale_casi, direction = input$direction_fill, method = input$method_fill)
})
output$coolplot1 <- plotly::renderPlotly({
wait <- waitLoading()
#waiter::waiter_show(id = "coolplot1", html = waiter::spin_loaders(id = 1, color = "#ff471a"), color = "white")
# Data trim and curve fitting #
if(is_ready(reac_general$sample_cases)) {
n <- nrow(eval(t$data)[[t$name]])
logic_interval <- reac_general$sample_date >= input$fitInterval[1] &
reac_general$sample_date <= input$fitInterval[2]
sample_serial_date <- eval(t$data)[[t$name]]$data_seriale
sample_cases <- reac_general$sample_cases
sample_diff <- c(NA,diff(sample_cases))
if( !t$p && input$swab_std ) {
swabs <- covid19:::level_and_fill(eval(t$data)[[t$name]]$tamponi, direction = input$direction_fill, method = input$method_fill)
sample_cases <- sample_cases / swabs
sample_diff <- c(NA,sample_diff[-1] / diff(swabs))
}
sample_serial_date_trim <- sample_serial_date[logic_interval]
sample_cases_trim <- sample_cases[logic_interval]
sample_diff_trim <- sample_diff[logic_interval]
sample_serial_date_rem <- sample_serial_date[!logic_interval]
sample_cases_rem <- sample_cases[!logic_interval]
sample_diff_rem <- sample_diff[!logic_interval]
fit_data <- covid19:::exe_fit(sample_cases = sample_cases_trim,
sample_date = sample_serial_date_trim,
days = days)
reac_logistic$model <- fit_data$out_fit$model
reac_logistic$resid <- fit_data$out_resid
reac_logistic$vals <- fit_data$out_fit$vals
conf <- nlstools::confint2(level = 0.95, object = reac_logistic$model)
yConf_up <- (conf["n0",2]*conf["k",2])/(conf["n0",2] + (conf["k",2]-conf["n0",2]) * exp(-conf["r",2]*(sample_serial_date_trim-sample_serial_date_trim[1]))) + sample_cases_trim[1]
yConf_down <- (conf["n0",1]*conf["k",1])/(conf["n0",1] + (conf["k",1]-conf["n0",1]) * exp(-conf["r",1]*(sample_serial_date_trim-sample_serial_date_trim[1]))) + sample_cases_trim[1]
# Conversion to real date and creation of fitted points #
points_trim <- data.frame("sample_serial_date_trim" = reac_general$sample_date[logic_interval],
sample_cases_trim)
points_rem <- data.frame("sample_serial_date_rem" = reac_general$sample_date[!logic_interval],
sample_cases_rem)
points_diff_trim <- data.frame("sample_serial_date_trim" = reac_general$sample_date[logic_interval],
sample_diff_trim)
points_diff_rem <- data.frame("sample_serial_date_rem" = reac_general$sample_date[!logic_interval],
sample_diff_rem)
fittedPoints <- fit_data$fittedPoints
fittedPoints_der <- fit_data$fittedPoints_der
seq_dates <- seq(from = init_date, by = 1, length.out = length(days))
fittedPoints$days <- seq_dates
fittedPoints_der$days <- seq_dates
confPoints_up <- data.frame("sample_serial_date_trim" = reac_general$sample_date[logic_interval],
yConf_up)
confPoints_down <- data.frame("sample_serial_date_trim" = reac_general$sample_date[logic_interval],
yConf_down)
# PLOT with plotly #
fig = plotly::plot_ly( name = "Cases", type= "scatter")
# funtions for the two different plots
plot1 = function(fig)
{
fig <- fig %>% plotly::add_trace(data = confPoints_down, x = ~sample_serial_date_trim, y = ~yConf_down, mode='none',hoverinfo='skip',fill = 'tozeroy', name="IGNORED_LEGEND", fillcolor="rgba(0,0,0,0)",showlegend = FALSE)
fig <- fig %>% plotly::add_trace(data = confPoints_up, x = ~sample_serial_date_trim, y = ~yConf_up, mode='none', fill = 'tonexty' ,name="Confidence interval 95%", fillcolor="rgb(255,250,205)")
hovlabels <- c("")
for(i in c(1:n)) {
hovlabels[i] <- paste(format(reac_general$sample_date[i], "%d %b"),
", Tot cases = ", sample_cases[i], sep ="")
if( !t$p && input$swab_std )
hovlabels[i] <- paste(hovlabels[i], ", Tot swabs = ", swabs[i], sep = "")
}
fig <- fig %>% plotly::add_trace(data = points_rem, x =~sample_serial_date_rem, y =~sample_cases_rem ,marker = list(color = "red"), mode = 'markers', name = "Total cases (excluded)",
text = hovlabels[!logic_interval], hoverinfo = 'text')
fig <- fig %>% plotly::add_trace(data = points_trim, x =~sample_serial_date_trim, y =~sample_cases_trim ,marker = list(color = "green"), mode = 'markers', name = "Total cases (fitting)",
text = hovlabels[logic_interval], hoverinfo = 'text')
fig <- fig %>% plotly::add_trace(data = fittedPoints, x = ~days, y = ~yFitted, line = list(color ='rgb(0,0,139)',width=2.5), mode='lines', name = "Fitted logistic curve" )
return(fig)
}
plot2 = function (fig)
{
hovlabels <- c("")
for(i in c(2:n)) {
hovlabels[i] <- paste(format(reac_general$sample_date[i], "%d %b"),
", Cases = ",
sample_cases[i] - sample_cases[i-1],
sep ="")
if( !t$p && input$swab_std )
hovlabels[i] <- paste(hovlabels[i], ", Swabs = ", swabs[i]-swabs[i-1], sep = "")
}
fig <- fig %>% plotly::add_bars(data = points_diff_rem, x =~sample_serial_date_rem, y =~sample_diff_rem, marker = list(color = "red"), name = "New cases (excluded)",
text = hovlabels[!logic_interval], hoverinfo = 'text')
fig <- fig %>% plotly::add_bars(data = points_diff_trim, x =~sample_serial_date_trim, y =~sample_diff_trim, marker = list(color = "green"), name = "New cases (fitting)",
text = hovlabels[logic_interval], hoverinfo = 'text')
fig <- fig %>% plotly::add_trace(data = fittedPoints_der, x = ~days, y = ~yFitted_der, line = list(color ='rgb(255,117,20)',width=2.5), mode='lines', name= "Fitted logistic distribution")
return(fig)
}
#Plot based on the checkbox
if( 1 %in% input$plot_type && !(2 %in% input$plot_type ) )
{
fig = plot1(fig)
} else if( 2 %in% input$plot_type && !(1 %in% input$plot_type ) )
{
fig = plot2(fig)
} else if( 1 %in% input$plot_type && (2 %in% input$plot_type ) )
{
fig = plot1(fig)
fig = plot2(fig)
}
# labels and plot
fig <- fig %>%plotly::layout(xaxis = list(title = "day"), yaxis = list(title = "Infected"))
if( reac_logistic$vals$k > 1e7 )
fig <- fig %>%plotly::layout(title = "Warning: unrealistic model estimated", font = list(color = 'red'),dtick= 5 )
fig
}
})
#-- Summary of regions ---
output$fit_smry <- shiny::renderPrint({
wait <- waitLoading()
summary(reac_logistic$model)
})
output$selected_terr <- shiny::renderPrint({
})
selected_info_func <- function() {
if(is_ready(t$name))
{
div(h2(strong(paste("Selected territory:", t$name))), align = "center", style = "color:red")
}
}
output$selected_info1 <- shiny::renderUI({
selected_info_func()
})
output$selected_info2 <- shiny::renderUI({
selected_info_func()
})
output$selected_info3 <- shiny::renderUI({
selected_info_func()
})
output$selected_info4 <- shiny::renderUI({
selected_info_func()
})
output$selected_info5 <- shiny::renderUI({
if(is_ready(reac_SEIR$name) && is_ready(reac_SEIR$notAvail))
{
if(reac_SEIR$notAvail) {
fluidPage(
div(h2(strong(paste("Selected territory:", reac_SEIR$name))), align = "center", style = "color:red"),
div(h5(em(paste("Data for SEIR not available at province level"))), align = "center", style = "color:red")
)
} else {
div(h2(strong(paste("Selected territory:", reac_SEIR$name))), align = "center", style = "color:red")
}
}
})
output$resid_smry <- shiny::renderPrint({
wait <- waitLoading()
print("Data: residuals from nls regression")
nlstools::test.nlsResiduals(reac_logistic$resid)
})
#-- Residuals ---
output$plot_residual <- plotly::renderPlotly({
wait <- waitLoading()
Res_DF_1<-as.data.frame(reac_logistic$resid$resi1)
Res_DF_2<-as.data.frame(reac_logistic$resid$resi2)
Res_DF_3<-as.data.frame(reac_logistic$resid$resi4)
Res_DF_4<-as.data.frame(reac_logistic$resid$resi3)
p = plotly::plot_ly(type = 'scatter')
p <- p %>% plotly::layout(
xaxis = list(
showline = FALSE,
dtick = 2000,
zeroline = FALSE
),
yaxis = list(
showline = FALSE,
zeroline = FALSE
)
)
if( is_ready(input$plot_res_type) ) {
if(input$plot_res_type=="Residuals"){
colnames(Res_DF_1)=c("fitted1","res")
p <- p %>% plotly::layout(
title ="Residuals",
xaxis = list(title="Fitted values",zeroline = FALSE),
yaxis = list(title="Residuals")
)
p= p %>%plotly::add_trace(name = "residual",showlegend=FALSE,data=Res_DF_1,x=~Res_DF_1$fitted1,y=~Res_DF_1$res,marker = list(size = 15,
color = 'rgba(255, 182, 193, .9)',
line = list(color = 'rgba(152, 0, 0, .8)',
width = 2)))
p <- p %>%plotly::add_trace(data=Res_DF_2,x=~Res_DF_1$fitted1,y = 0, name = "liney0", line = list(color = 'rgb(0,0,0)', width = 1, dash = 'dot'),showlegend = FALSE, mode = 'lines')
p
#grafico1
}
else if(input$plot_res_type=="Residuals_standardized"){
colnames(Res_DF_2)=c("fitted2","res_stand")
p <- p %>% plotly::layout(
title ="Residuals standardized",
xaxis = list(title="Fitted values"),
yaxis = list(title="Standardized residuals")
)
p <- p %>%plotly::add_trace(name="Standardised residuals",showlegend=FALSE,data=Res_DF_2,x=~Res_DF_2$fitted2,y=~Res_DF_2$res_stand,marker = list(size = 15,
color = 'rgba(255, 182, 193, .9)',
line = list(color = 'rgba(152, 0, 0, .8)',
width = 2)))
p <- p %>%plotly::add_trace(data=Res_DF_2,x=~Res_DF_2$fitted2,y = 0, name = "liney0", line = list(color = 'rgb(0,0,0)', width = 1, dash = 'dot'),showlegend = FALSE, mode = 'lines')
p <- p %>%plotly::add_trace(data=Res_DF_2,x=~Res_DF_2$fitted2,y = 2, name = 'liney2', line = list(color = 'rgb(0,0,0)', width = 1),showlegend = FALSE ,mode = 'lines')
p
}
else if(input$plot_res_type=="Autocorrelation"){
colnames(Res_DF_3)=c("fitted3","resiplus1")
p <- p %>% plotly::layout(
xaxis = list(title="Residuals i"),
yaxis = list(title="Residuals i+1"),
title ="Autocorrelation")
p = p %>%plotly::add_trace(name = "Autocorrelation", showlegend=FALSE,data=Res_DF_3,x=~Res_DF_3$fitted3,y=~Res_DF_3$resiplus1,marker = list(size = 15,
color = 'rgba(255, 182, 193, .9)',
line = list(color = 'rgba(152, 0, 0, .8)',
width = 2)))
p <- p %>%plotly::add_trace(data=Res_DF_3,x=~Res_DF_3$fitted3,y = 0, name = "liney0", line = list(color = 'rgb(0,0,0)', width = 1, dash = 'dot'),showlegend = FALSE, mode="lines")
p
#grafico1
}
else if(input$plot_res_type=="Sqrt of abs of res vs fitted"){
p <- p %>% plotly::layout(
title ="Sqrt of abs of residual",
xaxis = list(title="Fitted"),
yaxis = list(title="Residuals")
)
colnames(Res_DF_4)=c("fitted4","qq")
p = p %>%plotly::add_trace(name="Sqrt of abs of res vs fitted",showlegend=FALSE,data=Res_DF_4,x=~Res_DF_4$fitted4,y=~Res_DF_4$qq,marker = list(size = 15,
color = 'rgba(255, 182, 193, .9)',
line = list(color = 'rgba(152, 0, 0, .8)',
width = 2)))
p
#grafico1
}
}
})
##===================================== ARIMA SECTION =============================================##
## HERE GOES ALL ARIMA IMPLEMENTATION COMMON TO ALL GRAPHS
shiny::observe({
wait <- waitLoading()
if(is_ready(input$arima_interval[1]) && is_ready(reac_general$sample_cases)) {
reac_ARIMA$logic_interval <- reac_general$sample_date >= input$arima_interval[1] &
reac_general$sample_date <= input$arima_interval[2]
reac_ARIMA$sample_date_trim <- reac_general$sample_date[reac_ARIMA$logic_interval]
reac_ARIMA$sample_cases_trim <- reac_general$sample_cases[reac_ARIMA$logic_interval]+1
reac_ARIMA$sugg_lags <- suggest_lags()
updateSliderInput(session,"ARIMA_p", value= reac_ARIMA$sugg_lags[1])
updateSliderInput(session,"ARIMA_I", value = reac_ARIMA$sugg_lags[2])
updateSliderInput(session,"ARIMA_q", value = reac_ARIMA$sugg_lags[3])
}
})
shiny::observe({
wait <- waitLoading()
if(is_ready(input$ARIMA_p)) {
reac_ARIMA$arima <- checkExp( stats::arima(log(reac_ARIMA$sample_cases_trim),order=c(input$ARIMA_p,input$ARIMA_I,input$ARIMA_q)) , "There is not a suitable ARIMA model")
reac_ARIMA$arimaOK <- TRUE
}
})
#Ln x = Log10 x / Log10 e
## Plot of ARIMA Time Series and FORECAST
output$arima_coolplot1 <- plotly::renderPlotly({
wait <- waitLoading()
#acf(log(sample_cases_trim))
#pacf(log(sample_cases_trim))
if( reac_ARIMA$arimaOK ) {
fore = checkExp(forecast::forecast(reac_ARIMA$arima,input$forecast) , "There is not a forecast for the ARIMA model")
sdt <- reac_ARIMA$sample_date_trim
fore.dates <- seq(from = sdt[length(sdt)], by = 1, len = input$forecast)
p <-plotly::plot_ly() %>%
plotly::add_ribbons(x = fore.dates,
ymin = exp(fore$mean),
ymax = exp(fore$upper[, 2]),
color = I("#17becf"),
name = "95% confidence") %>%
plotly::add_ribbons(p,
x = fore.dates,
ymin = exp(fore$mean),
ymax = exp(fore$upper[, 1]),
color = I("#ed9dac"), name = "80% confidence")%>%
plotly::add_lines(x = sdt, y = reac_ARIMA$sample_cases_trim,
color = I("#037d50"),
name = "observed",
mode="lines")%>%
plotly::add_lines(x = fore.dates,
y = exp(fore$mean),
color = I("#ee1147"),
name = "prediction")
p <- p %>% plotly::layout(
title = paste0("ARIMA Forecast (",input$ARIMA_p,",",input$ARIMA_I,",",input$ARIMA_q,")"),
xaxis = list(title="Days"),
yaxis = list(title="Total cases", type = "log")
)
p
}
})
## Plot of autocorrelation function
output$arima_coolplot2 <- plotly::renderPlotly({
wait <- waitLoading()
if(is_ready(reac_ARIMA$sample_cases_trim)) {
p = checkExp( forecast::autoplot(acf(log(reac_ARIMA$sample_cases_trim))) , "There is not a forecast for the ARIMA model")
p = plotly::ggplotly(p)
p <- p %>%plotly::layout(xaxis = list(title = "LAG"), yaxis = list(title = "ACF"), title="Autocorrelation", showlegend = TRUE,
plot_bgcolor = "rgb(255, 255, 255)")
}
})
## Plot of partial autocorrelation function
output$arima_coolplot3 <- plotly::renderPlotly({
wait <- waitLoading()
if(is_ready(reac_ARIMA$sample_cases_trim)) {
p = checkExp( forecast::autoplot(pacf(log(reac_ARIMA$sample_cases_trim)) ) , "There is not a forecast for the ARIMA model")
p = plotly::ggplotly(p)
p <- p %>%plotly::layout(xaxis = list(title = "LAG"), yaxis = list(title = "PACF"), title="Partial Autocorrelation", showlegend = TRUE,
plot_bgcolor = "rgb(255, 255, 255)")
}
})
## Plot of ARIMA residuals
output$arima_coolplot4 <- shiny::renderPlot({
wait <- waitLoading()
if(reac_ARIMA$arimaOK) {
checkExp(forecast::checkresiduals(reac_ARIMA$arima) , "There is not a suitable ARIMA model")
}
})
# --- Summary ARIMA
## Print of suggested parameters
output$parameters_sugg <- shiny::renderUI({
wait <- waitLoading()
if( is_ready(reac_ARIMA$sugg_lags[1]) ) {
fluidPage(
fluidRow(
h3( paste("Suggested Parameters: ARIMA(", paste(reac_ARIMA$sugg_lags, collapse = ","), ")") )
),
fluidRow(
h4(paste("Suggested initial and final date: ", reac_general$sugg_dates[1]," - ", reac_general$sugg_dates[2]))
)
)
}
})
# toString(reac_ARIMA$arima$coef)
# suggested fit toString(forecast::auto.arima(log(sample_cases_trim)))
#reac_ARIMA$arima$coef
# reac_ARIMA$arima$sigma2
output$arima_shell_output <- shiny::renderPrint({
wait <- waitLoading()
if(reac_ARIMA$arimaOK) {
reac_ARIMA$arima
}
})
output$arima_shell_resid <- shiny::renderPrint({
wait <- waitLoading()
if(reac_ARIMA$arimaOK) {
checkExp(forecast::checkresiduals(reac_ARIMA$arima,plot=FALSE), "There is not a suitable ARIMA model")
}
})
#============ FFT ======================================
shiny::observe({
wait <- waitLoading()
if(is_ready(reac_general$sample_cases)) {
reac_FFT$logic_interval <- reac_general$sample_date >= input$FFT_interval[1] &
reac_general$sample_date <= input$FFT_interval[2]
reac_FFT$sample_date_trim <- reac_general$sample_date[reac_FFT$logic_interval]
reac_FFT$sample_cases_trim <- reac_general$sample_cases[reac_FFT$logic_interval]+1
}
})
output$FFT_day_cases<- shiny::renderPlot({
if(is_ready(reac_FFT$sample_cases_trim)) {
FFTX<-spectral::spec.fft(diff(reac_FFT$sample_cases_trim))
plot(FFTX,type = "l",ylab = "Amplitude",xlab = "Frequency",lwd = 2)
}
})
output$FFT_day_cases_diff<- shiny::renderPlot({
if(is_ready(reac_FFT$sample_cases_trim)) {
FFTX<-spectral::spec.fft(diff(diff(reac_FFT$sample_cases_trim)))
plot(FFTX,type = "l",ylab = "Amplitude",xlab = "Frequency",lwd = 2)
}
})
#======================= R(T) ==================================
#DEFAULT 1.4, 0.8
#
#
#shiny::observe({
#
# if(is_ready(reac_general$sample_cases) && is_ready(input$"Gamma_1")) {
#
# GT.chld.hsld2<-R0::generation.time("gamma", c(input$"Gamma_1", input$"Gamma_2"))
# time_R <- as.numeric(length(reac_general$sample_cases))-2
#
# tryCatch(
# {
# reac_R$R0_data <- R0::est.R0.TD(diff(reac_general$sample_cases),GT.chld.hsld2, begin=1, end=time_R)
# },
# error = function(e)
# {
# reac_R$R0_data <- FALSE
# }
# )
# }
#})
#
#R0_data <- shiny::reactive({
#
# shiny::validate(
# shiny::need(is_ready(reac_R$R0_data), "Wait...") %then%
# shiny::need(!isFALSE(reac_R$R0_data), "The estimates cannot be calculated")
# )
# return(reac_R$R0_data)
#})
#
#
#
#output$R_t_evaluation<- shiny::renderPlot({
#
# plot( R0_data())
#
#})
#
#output$R_t_goodness_of_fit<- shiny::renderPlot({
#
# R0::plotfit(R0_data())
#
#})
#
#output$R_t_evaluation_FFT<- shiny::renderPlot({
#
# R0_data_raw_FFT <- spectral::spec.fft(R0_data()[["R"]])
# plot(R0_data_raw_FFT,type = "l",ylab = "Amplitude",xlab = "Frequency",lwd = 2)
#
#})
#======================= SEIR MODEL ==================================
#
#
# #---- data retrieval, common to both
#
# distr_IT <- 'exp'
# distr_SRT <- 'exp'
# distr_IBST <- 'none'
#
# shiny::observe({
#
# wait <- waitLoading()
#
# # territory selection for SEIR
# if(t$p) {
# reac_SEIR$data <- expression(regionTS)
# reac_SEIR$notAvail <- TRUE
# reac_SEIR$name <- regAndProv[ regAndProv$province == t$name, "region"]
# } else {
# reac_SEIR$data <- t$data
# reac_SEIR$notAvail <- FALSE
# reac_SEIR$name <- t$name
# }
#
# # N = population of area
# if(t$c) {
# reac_SEIR$N <- italy_pop$country[1,"valore"]
# } else {
# reac_SEIR$N <- italy_pop$region[italy_pop$region$territorio == reac_SEIR$name, "valore"]
# }
#
# ## Time series
# # Infected Time Series
# reac_SEIR$P <- level_and_fill(eval(reac_SEIR$data)[[reac_SEIR$name]]$totale_casi, direction = input$direction_fill, method = input$method_fill)
#
# # Removed Time Series (recovered + dead)
# reac_SEIR$R <- level_and_fill(eval(reac_SEIR$data)[[reac_SEIR$name]]$dimessi_guariti, direction = input$direction_fill, method = input$method_fill) +
# level_and_fill(eval(reac_SEIR$data)[[reac_SEIR$name]]$deceduti, direction = input$direction_fill, method = input$method_fill)
#
#
# reac_SEIR$par_IT <- list('rate'=1/input$rate_IT)
# reac_SEIR$par_SRT <- list('rate'=1/input$rate_SRT)
# reac_SEIR$par_IBST <- NULL
#
# reac_SEIR$IT <- covid19::generate(distr_IT, reac_SEIR$par_IT)
# reac_SEIR$SRT <- covid19::generate(distr_SRT, reac_SEIR$par_SRT)
# reac_SEIR$IBST <- covid19::generate(distr_IBST, reac_SEIR$par_IBST)
#
# reac_SEIR$m <- max(length(reac_SEIR$IT), length(reac_SEIR$SRT), length(reac_SEIR$IBST))
# reac_SEIR$end <- length(reac_SEIR$P) #Rt goes from day 1 to day n_obs-m BUT also the SEIR data series will go from day 1 to that day after refinement
# reac_SEIR$len <- reac_SEIR$end - reac_SEIR$m
# })
#
#
# #---- R0 part
#
# shiny::observeEvent(input$est_R0, {
#
# out <- SEIR_factotum(P = reac_SEIR$P, R = reac_SEIR$R, N = reac_SEIR$N,
# end=reac_SEIR$end, future=input$future,
# distr_IT=distr_IT, distr_SRT=distr_SRT, distr_IBST=distr_IBST,
# par_IT=reac_SEIR$par_IT, par_SRT=reac_SEIR$par_SRT, par_IBST=reac_SEIR$par_IBST,
# R0_exp_est_end=input$R0_exp_est_end,
# plot_data=input$plot_data)
#
# reac_SEIR$R0 <- out$R0
#
# S_ <- reac_SEIR$N*out$S_
# E_ <- reac_SEIR$N*out$E_
# I_ <- reac_SEIR$N*out$I_
# R_ <- reac_SEIR$N*out$R_
# S <- reac_SEIR$N*out$S
# E <- reac_SEIR$N*out$E
# I <- reac_SEIR$N*out$I
# R <- reac_SEIR$N*out$R
#
# reac_SEIR$U <- data.frame(date = seq(from = init_date, by = 1, length.out = reac_SEIR$len), 'S'=S, 'E'=E, 'I'=I, 'R'=R)
# reac_SEIR$U_ <- data.frame(date = seq(from = init_date, by = 1, length.out = reac_SEIR$len), 'S_'=S_, 'E_'=E_, 'I_'=I_, 'R_'=R_)
#
# reac_SEIR$current <- "R0"
#
# })
#
#
# shiny::observeEvent(input$est_Rt, {
#
#
# # REFERS TO INPUT IN PRECEDENT TAB
#
# GT<-R0::generation.time("gamma", c(input$"Gamma_1", input$"Gamma_2"))
# R0_out <- R0::est.R0.TD(diff(reac_SEIR$P),GT,begin=1, end=as.numeric(reac_SEIR$len))
# Rt <- R0_out$R
# R0_stages <- (1:(length(Rt)-1))
#
# out <- SEIR_factotum(P = reac_SEIR$P, R = reac_SEIR$R, N = reac_SEIR$N,
# end=reac_SEIR$end, future=input$future,
# distr_IT=distr_IT, distr_SRT=distr_SRT, distr_IBST=distr_IBST,
# par_IT=reac_SEIR$par_IT, par_SRT=reac_SEIR$par_SRT, par_IBST=reac_SEIR$par_IBST,
# R0_msp=Rt, R0_stages=R0_stages,
# plot_data=input$plot_data)
#
# S_ <- reac_SEIR$N*out$S_
# E_ <- reac_SEIR$N*out$E_
# I_ <- reac_SEIR$N*out$I_
# R_ <- reac_SEIR$N*out$R_
# S <- reac_SEIR$N*out$S
# E <- reac_SEIR$N*out$E
# I <- reac_SEIR$N*out$I
# R <- reac_SEIR$N*out$R
#
# reac_SEIR$U <- data.frame(date = seq(from = init_date, by = 1, length.out = reac_SEIR$len), 'S'=S, 'E'=E, 'I'=I, 'R'=R)
# reac_SEIR$U_ <- data.frame(date = seq(from = init_date, by = 1, length.out = reac_SEIR$len), 'S_'=S_, 'E_'=E_, 'I_'=I_, 'R_'=R_)
#
# reac_SEIR$current <- "R(t)"
#
# })
#
#
# shiny::observe({
# if(is_ready(reac_SEIR$U_) ) {
# reac_SEIR$SEIR_plot <- highcharter::hchart(tidyr::gather((reac_SEIR$U_), key="key", value="value", -date),
# highcharter::hcaes(x = date, y = value, group = key),
# type = "line",
# color = c("#bf40bf", "#e60000", "#00b300", "#fec501"),
# name = c("Exposed", "Infectious", "Recovered", "Susceptible"),
# marker = list(enabled = FALSE),
# legendIndex = c(2,3,4,1),
# tooltip = list(valueDecimals = 0)
# ) %>%
# highcharter::hc_add_series(tidyr::gather((reac_SEIR$U), key="key", value="value", -date),
# highcharter::hcaes(x = date, y = value, group = key),
# type = "scatter",
# name = c("Exposed (true)", "Infectious (true)", "Recovered (true)", "Susceptible (true)"),
# color = c("#bf40bf", "#e60000", "#00b300", "#fec501"),
# legendIndex = c(2,3,4,1),
# tooltip = list(valueDecimals = 0,
# pointFormat = '<span style="color:{point.color}">■</span> {series.name}: <b>{point.y}</b><br/>',
# headerFormat = '<span style="font-size: 10px">{point.key}</span><br/>')
# ) %>%
# highcharter::hc_title(text = paste("SEIR based on", reac_SEIR$current))
# }
# })
#
#
# output$SEIR_R0 <- renderPrint({
# if(is_ready(reac_SEIR$R0)) {
# print(reac_SEIR$R0)
# }
# })
#
# output$SEIR_plot <- highcharter::renderHighchart(
# reac_SEIR$SEIR_plot
# )
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
Embedding an R snippet on your website
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.