miraisolutions / covid19 Goto Github PK

There are a few problems with the area plots by status. The most relevant ones are in bold.

• The x axis is wrong for both the stacked areaplot and the top-country linepolot , probably shifted
  

The way we work around the issue with stacked log-scale by plotting the dodged cumulative sum of the data has some unwanted consequences

• Since the areas are transparent, the colors are mixed in => this could be fixed easily with a workaround
• Since the plot is interactive, on mouse over it shows the cumulative sum, not the actual number => must be fixed. Related to this, if one de-selects one of the low series, their area is then shown has belonging to the higher series
• the plot looks pretty heavy, especially since plotly does not do a very good job with the areas since it draws a border around the whole area
• There is still something wrong with the handling of zeros in the log scale, ~~although the issue once we have the active as first plot (can be covered in this issue), although the approach is anyhow not very robust => approach slightly improved, note howeverthe order of the stacked areas was not sensible when always starting with active, and we actually at the present date. A general improvement could be to switching to step plots, see #35

At the moment we have 2 sets of boxes one on top of eachother
Say

---------------------------- box1
Confirmed
100'000
---------------------------- box2
New
500
-----------------------------

It was a problem to do only one box due to CSS and we went temporarily for this sub-optimal solution
We would like instead to have 1 box:

---------------------------- 
Confirmed: 100'000
New: 500
------------------------

or

---------------------------- 
Confirmed           New
100'000               500
------------------------

legend should be the same for an indicator across all dates, but yet able to show relevant differences across countries.

develop ui and server for module country_comparison

font for value in valueBox module global is very small.

add Mirai Logo, copyright and contacts (e.g. with a footer) and colors

when clicking on the map, the pop up should show more information, namely:

• Country
• Confirmed
• Deaths
• Active
• Recovered
• (Growth rate)
• (Death rate)

Make the fetching of the data more efficient

add option to radio button to show the new confirmed cases per day

If you select 3 countries you will see that country colours are changing in the different plots:n
Say in lineplot over dates:
Italy Red
Switzerland blue
France green

In plot from day of 1000 contagions
Italy green
Switzerland blue
France red

Make sure the countries have the same colours in all the line plots

Remove pop-up disclaimer and, instead, add the same information in the header.
Preferable solution: A pop-up message whenmousing over the subtitle.

to show that something is being loaded while the user waits

• fonts are too big and we must find a way to say 10000, like 10k, to adjust also the names in X and Y (to be checked in other sites)

• more space between first row of bar plots and second row of bar plots in the country tab

• lines in plot are perhaps too thick --> related to plotly

• Date can be removed from x axis in plots

• mouse over does not work too well, also too many options --> related to plotly

• improve colors

• fix add polygons update
• fix map height

Add text indicating where the data comes from either on the Global tab or in the header

table with count, it should start from today in first row by default

possibility to switch between the two scales in the plot

• logo with inverted colors in header top right
• add version to app title reading it from DESCRIPTION file
• in the header "Covid19, v** by Mirai Solutions GmbH" with Mirai Solutions GmbH linked to website
• add widget to show what is new compared to previous version => see #43
• add CHANGELOG NEWS.MD file to repository

Requires info of countries population

Adding at least in the table on the global tab the variable confirmed x 100m inhabitants.
The world population by country can be found in the following link. It's a good dataset because it has country codes as well as country names.

https://data.worldbank.org/indicator/SP.POP.TOTL
api.worldbank.org/v2/en/indicator/SP.POP.TOTL?downloadformat=csv

requires knowing which countries belong to which continent.

if the legend contains only the one country being plotted, then hide legend to save space. valid for module country

values in the Y axes are wrong, they are all the same for all the histograms

Country labels in barplots are overlapping in most of the devices, only with large screen they would display correctly.
Add an angle of 30 degrees to prevent overlap

automatic deployment is currently failing https://travis-ci.com/github/miraisolutions/Covid19
when publishing from RStudio, what is rendered is the README https://github.com/miraisolutions/Covid19/

The count boxes with number of cases by type relies on shinydashboard, which is not worth having as dependencies just for this, plus the rendering of the boxes is not optimal.

The boxes should be rendered using customized divs and modularized for usage in both the global and country tabs.

we want users to be able to see what has changed among versions.

This most likely implies to also include the date in the NEWS.md alongside the the version (e.g. # Covid19 1.0.0 (2020-03-XX))

In the stacked line plot smarp style we have from bottom to top

• deaths
• recovered
• active

Total makes confirmed.
In the country page the order is different, this is actually a bug.

Invert order to show better the curve of active, the downward trend.

From bottom to top:

• active
• deaths
• recovered

In the world map if I check "Confirmed" cases I have a color scale from 1 to 10 millions.
If then I click on "per 1M pop" then I still have scale from 1 to 10 millions.
i.e. max 10 million cases over a million population.
The colors seem in fact much lighter.
The scale should go possibly from 1 to 10000 (guess)

The same happens for all the variables, in particular those "New".

To check also if perhaps we are still considering small countries like San Marino that would mess up the color scale

add plot in global tab comparing countries per day from the day of the 100th contagion

https://github.com/CSSEGISandData/COVID-19/tree/master/csse_covid_19_data/csse_covid_19_time_series#readme

use renderPlotly instead of renderPlot

tab with info for Europe only

Requires info of which countries are Europe and which are not

Use Thousand separator in Boxes and insert NEW cases in boxes

Add map of the world showing active, deaths, recovered and confirmed cases

https://www.worldometers.info/world-population/population-by-country/

LIneplot from day of 1000 contagion

For some variables are the lines do not look too good. Say attached pic:
looking bad: New Confirmed, New Active
partially affected: growth factor, lethality rate
looking ok: the rest

If doing a rolling average for only 2 variables could create problems then it can be done for all.

Rolling average in day X: mean(Day(x)..... Day(x-7))

Remove also dots and keep only line plot.
Despite I like the dots initially now that we have more days they make the plot harder to read.
Alternative could be to have a dot every 7 days

after the implementation of #56 we have now new active as an option for the map. This can have negative values, currently outside of our legend range. Fix this by adding a color for negative values.

Consider rendering the area plots with steps (via geom_crossbar()) as opposed to the current linear (viageom_ribbon(), as this reflects better the nature of the data w/o linear interpolation artifacts.

There is a commented out usage of geom_crossbar() in 69db01d

Current top 5 countries growth factor
Current top 5 countries death toll
Top 5 countries from day of 1000 contagion

The titles above appear also in the comparison tab, change it (only here) to

Growth factor
Death toll
Comparison from day of 1000 contagion

e.g. death rate

new cases should be computed from the data. Notice that there are different interpretation of new cases:

• tot_cases_today - tot_cases_yesterday
• tot_cases_today - tot_cases_yesterday - new_deaths - new_recovered

In global page and comparison page we can add a scatterplot:

X axis: Confirmed cases over one million
Y axis: Growth factor (based on confirmed case)

how plot should look like:

• dot on value: country name (text) above value

if achievable, colors to dots and texts:

• color countries on top right with red (bad)
• color countries on bottom left with green (good)
• color countries on bottom right/top left with yellow (not that ok , not that bad)

to define and divide the plot into 4 rectangles, options:
a) use the countries selected in the graph, take median X and median Y
b) take all countries with >1000 cases, take median X median Y

confirmed = red
deaths = black
recovered = green
active = blue

when showing confirmed, active, deaths and recovered in one plot, use area plots instead of line plots

swiss data available here https://github.com/openZH/covid_19
for the time being it is only about the new cases, no info about deaths or recovered

currently I plot the log of the value instead of using logaritmic scale. to be fixed

remove all the search box in the first page, they do not search much, only filter making sense is with dates

Controlling the package dependencies and R version is crucial to a smooth collaborative development and safe deployments. See also #116 for a recent breakage caused by picking the latest shiny 1.5.0

• R version: Note that shinyapps.io serves the app using the R version used when running rsconnect::deployApp(), so we are not bound to the latest R version and can set it explicitly in the setup-r action as r-versionR 4.0.0 is out, and shinyapps.io will be updated accordingly soon (expected by around May 3rd, see https://community.rstudio.com/t/shinyapps-io-cant-deploy-apps-built-with-r-4-0-0/63143/3)

• Package dependencies: We can use full-control via renv (as in https://github.com/miraisolutions/eRum2020Workshop) and / or rely on an MRAN / RSPM snapshot => renv setup as part of #151

As part of this, we should also review the CI-CD Actions workflow WF and setup following https://github.com/miraisolutions/eRum2020Workshop (which includes making sure the cache is Rmajor.minor-specific), and adapt it to the agreed package dependencies control.