the-strategy-unit / 723_mh_covid_surge_modelling Goto Github PK

Connect internal subpopulation groups or switch to treemap plot?

The sum of the conditions sliders should not exceed 100%: currently this is enforced by setting the max-value of each slider independently to enforce this limit.

Instead, we should set the max-value's to 100% and use the following algorithm:

1. if the percent sum exceeds 100% find the minimum percentage of the other sliders (the ones that haven't been updated)
2. calculate the amount we need to reduce the sum by to reach 100%, and then equally divide this value by the number of sliders we are working with
3. take the minimum value of 1 and 2, then reduce all of the sliders by this amount
4. if the sum still exceeds 100%, remove the sliders which are now at 0% and repeat

Month is displaying as NaN in tooltips

add a tabItem/menuItem to the UI for a summary of surge demand by condition (as per page 12 of report)

• add a plot to show this data
• add a summary table (filtered to just show the months)

If the first treatment option is renamed and changed the values aren't correctly updated when a file is uploaded. A temporary fix is to re-run the upload a second time

steps to recreate:

1. start app
2. switch to results tab
3. hit download report button

There should be a input box for the size of each population subgroup, as well as an input box for the % of that population subgroup that we are using in the model. E.g. population a = 100,000, we model 10% of this group = 10,000

If you update any of the treatment sliders, only the currently selected population group's model is re-run, so the values are not correctly updated.

At the moment the only way to alter the demand data is to download parameters and update the excel file, then upload this new file.

It would be useful to be able to directly enter this data in the app

• add in a page to upload the service profile + suppressed demand data
• add this data into a plot on the "Results" page

condition pcnt slider is using the treatment pcnt figure instead

as this is no longer a "concept" it probably should be renamed

1. go to results tab, select a different treatment
2. go to params tab, alter some of the treatment params
3. go back to results tab, it is now back on the first treatment

now, if you instead

1. go to results tab, select a different treatment
2. go to params tab, do nothing
3. go back to results tab, it is still on the selected treatment

the app should behave consistently like in the second scenario

parameter download is not currently working, it is downloading a html document that doesn't appear to contain any of the param data

Each population subgroup should be able to have a different scenario selected for it. Currently you can only select a single scenario for every population subgroup

• separate the different parameter sections into

  • population groups
  • groups to conditions (g2c)
  • conditions to treatments (c2t)
  • treatments

• separate the g2c and c2t percentages into individual sliders

• add rule to ensure that the g2c percentages for that group sums to less than 1

• add rule to ensure that the c2t percentages for that condition sums to less than 1

• add the currently hidden treatment parameters (months, decay) to the "demand" section

• move the "Success % of treatment" to the "demand"

• rename the "demand" section to "treatment"

allow a user to upload a params.json file that they have previously downloaded and replace the currently loaded params item.

Will need to check that the params is valid (e.g. is of a similar structure to the existing dataset)

• Add a front page that explains the model and gives instructions on use
• include a diagram of the SD model

• Add an ordered (horizontal) bar chart to the results page that breaks down which population groups contribute to the "Total 'surge' referrals" output box.
• Show the values of the bars as labels at the end of each bar

Create a visual of the population groups at baseline, colours by their higher level descriptions (page 10 of report). Suggestion of visualisation is to use a circle packing method

add a tabItem/menuItem to the UI for a summary of surge demand by condition (as per page 13 of report)

• add a plot to show this data
• add a summary table (filtered to just show the months)

The number of months input box allows extreme values, e.g. 230, which can take a long time to model. Limits should be set (e.g. 0 to 23)

We could use javascript to save the currently set param's used, and then reload these when a user comes back to the app. This would save them having to download/upload params.

Fix NSE reference of variable, or find alternative solution

The first column shows "title" instead of the correct text for the column heading

• remove the old condition pcnt slider and its corresponding server code
• move the condition drop down to the "People being treated of condition group" panel
• remove the treatment drop down in the "People being treated of condition group" panel
• dynamically create each treatment in the "People being treated of condition group" (as per the condition panel)
• each treatment should have an numeric input box for the "proportion" of condition group and a slider for the treat %
• the "proportion" boxes should then be used to work out the c2t percentage by dividing each value by the sum: we could possibly try to show this percentage next to the numeric input to give the user an idea of what percentages they are setting

• add unit tests to the package
• set up code coverage with codecov.io
• set up an action to run code coverage
• ensure 100% coverage

add a tabItem/menuItem to the UI for a summary of surge demand by condition (as per page 14 of report)

• add a plot to show this data
• add a summary table (filtered to just show the months)