updating to directly input ACH and infer riskiness. User defined efficacy from delta ACH by cwhittaker1000 · Pull Request #142 · mrc-ide/helios

cwhittaker1000 · 2026-01-22T00:12:02Z

As above; this addresses and answers #136 - @geethaj1 will be handling this.

cwhittaker1000 · 2026-01-22T00:24:45Z

Aim with this PR is get helios into a position where we're directly inputting ACH values (probably specifying a distribution that we draw from) into the run_function, instead of riskiness parameters (specifying a riskiness distribution, itself derived externally from ACH values, see here).

The aim with doing this is to enable us to calculate efficacy not as a single input value, but as a function of ACH in each location; whilst also retaining the ability to specify riskiness, again in a location specific manner.

This is very rough, but based on our convo today, it seems like there are 4 broad pieces of work to be done here (not necessarily in this order):

Change the arguments and inputs of run_simulation (and the associated functions for generating the parameter list) to take in parameters that specify ACH values, rather than riskiness values. For example, these might be the setting-specific mean and sd of a normal distribution (but let's chat more about the exact parameterisation).
Change the arguments and inputs of run_simulation to take in parameters that specify the relationship between ACH and efficacy, conditional on e.g. some max_efficacy (or similar, we can talk about the exact parameterisation).
Use those parameters inside run_simulation (possibly within their own function, which is called within run_simulation) to draw an ACH for each location (e.g. for each location of a setting type, draw from a normal with setting-type specific ACH mean and sd defined in 1))
Use those ACHs for each location to get a corresponding riskiness (probably using the framework we initially developed in
this).
Use those ACHs to also calculate for each location (of each setting type) an intervention efficacy that is calculated based on their ACH and the new parameter inputs specified in 1).

…ose on the riskiness distribution

… riskiness.

geethaj1 · 2026-02-11T21:24:11Z

Updates:

Code to explore normal and lognormal distributions of ACH and impact on riskiness distributions in R/normal_dist_ACH_exploration.R and R/lognormal_ACH_dist.R
Created script r/setting_specific_ACH.R which includes:

function to draw ACH values from truncated normal distribution generate_setting_specific_ach()
function to convert raw ACH to riskiness value convert_ach_to_riskiness(), using the W-R model and parameters
placeholder function to calculating efficacy from ach calculate_efficacy_from_ach() (function not defined yet)
function for user to specify ACH distribution parameters set_setting_specific_ach()

Modified parameters.R

added W_R model parameters (emission rate, decay rate, p_inf, etc)
added setting specific room volumes
added ACH distribution parameters (mean, sd) for each setting type
added efficacy parameters (placeholders, haven't defined function)

Modified variables.R
-replaced riskiness generation with ACH

(placeholder) location specific efficacies based on ACH
I

… for relationship

…r each type of location, etc.)

…etting_specific_ach script

geethaj1 · 2026-02-25T23:29:35Z

Summary of changes:

set_setting_specific_riskiness.R is replaced by set_setting_specific_ach.R where user specifies the parameters of ach truncated normal distributions for each setting type instead of specifying the parameters of riskiness truncated lognormal distributions through user-facing set_setting_specific_ach() function.
ACH values are converted to riskiness using W-R through the convert_ach_to_riskiness()
ACH values are also used to calculate efficacy, where the user can choose from a constant or sigmoid function to define the relationship between ach and efficacy of intervention. The user does this through the set_uvc_ach() function

… estimates

…ific function based on tests

… tests to check that it works

…tions to interventions.R

cwhittaker1000 · 2026-05-19T17:48:56Z

@geethaj1 - quick q as I'm running through the code. Are there any files in inst that directly relate to or inform the code/default parameters in the model? If so, I'm thinking we should make a separate folder in inst called VIP or similar where we keep track of all these scripts. I'm motivated by the fact I keep trying to find that lognormal params script I wrote ages ago in inst whenever we want to look something related to that up. Would be good to have them all in one place - do you have any scripts like that or are they all mainly just for testing the model after you've made changes?

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>

cwhittaker1000

Great stuff @geethaj1 - really impressive work here. Everything looks broadly right to me - some minor comments/questions enclosed and I think the big piece is tidying this up and documentation improvements ahead of merge. Congrats!

cwhittaker1000 · 2026-05-19T17:53:24Z

+    volume_per_person_leisure = 8, # 2m^2*4M
+    volume_per_person_household = 50, # 20m^2 * 2.5m
+
+    # W-R parameters


Let's have a conversation about how much we want to expose these parameters - I think doing so is fine but useful to have a conversation about helper functions vs forcing users to input these directly etc.

cwhittaker1000 · 2026-05-19T17:55:41Z

+  #   far_uvc_household_sigmoid_x0 = NULL,
+
+  # Intervention parameters (Wells-Riley ACH-based efficacy):
+    intervention_joint_active              = FALSE,


I don't think any of these parameters are currently documented in the above documentation on inputs or elsewhere? I'm currently having to infer and understand them based on context and reading other parts of the codebase, which is time consuming. Can you make sure all these are documented and described please?

Added documentation for volume per person parameters, Wells Riley parameters, and intervention parameters

cwhittaker1000 · 2026-05-19T17:57:26Z

+    intervention_household_covered         = NULL,

    # Room Size Per Individual Parameters:
    size_per_individual_workplace = 1,


Do we still need these if we have the volume parameters above?

the volumes are used in the W-R calculations, but right now room size per individual is being used in the coverage allocation when coverage_target = "square_footage". We could change it so we change the square-footage option to volume, and then have coverage allocated by volume, which could be good to do

cwhittaker1000 · 2026-05-19T18:31:30Z

+  # calculation runs through the same code path.
+  for (s in c("workplace", "school", "leisure")) {
+    parameters_list[[paste0("intervention_", s, "_active")]]   <- TRUE
+    parameters_list[[paste0("intervention_", s, "_list")]]     <- parameters_list$intervention_joint_list


What is this parameter intervention_joint_list - I can't figure it out and where it's used/what for?

Is the list of things like what it does to modify ACH etc that we discussed?

Yes, intervention_joint_list stores the list of interventions (in this branch its just 1 intervention), and describes the characteristics of the intervention. It's used in calculate_efficacy_from_ach to get information about the interventino, and in generate_joint-intervention_switches for coverage allocation.

cwhittaker1000 · 2026-05-19T18:38:15Z

+                              variation_function       = NULL,
+                              variation_params         = list(),
+                              coverage                 = NULL) {
+  list(


I think we need some checks in here - if affected_by_baseline_ach is set to FALSE, then baseline_ach_function should be NULL right?

Related point is that none of these parameters are documented or particularly clear - I think we need clearer documentation of what they are and what they mean/correspond to.

ah sorry this is poor naming on my end, I've renamed affected_by_baseline_ach todelta_depends_on_baseline_ach delta_depends_on_baseline_ach , and baseline_ach_function to delta_function, as it is the function to define the delta ACH of an intervention, irrespective of whether it is affected by the baseline .
I've added added roxygen documenting the definitions

cwhittaker1000 · 2026-05-19T18:39:12Z

+                                 coverage_target,
+                                 coverage_type,
+                                 timestep,
+                                 ...) {


Instead of having the ... why not just force the user to pass in the list directly?

have edited to force user to pass directly, this was leftover from when I was initially attempting multiple interventions on this branch. Has now been cleaned up and documented

cwhittaker1000 · 2026-05-19T18:40:10Z

+# Efficacy at location i = 1 - p_post[i] / p_pre[i], where the post-intervention
+# alpha is augmented by the sum of intervention deltas (zeroed for uncovered
+# locations via the coverage vector).
+calculate_efficacy_from_ach <- function(ach_values, parameters_list, setting) {


this is really nice work Geetha!

cwhittaker1000 · 2026-05-19T18:40:55Z

+
+
+# =============================================================================
+# Helper functions for ACH / efficacy / UV-C conversions


haven't looked at the below in detail as presume they're helper functions not used anywhere much. Let me know if that seems wrong

this is right, the helper functions are just there to help the user figure out the appropriate inputs, not used anywhere in the actual model

baseline_ach_function -> delta_function baseline_ach_params -> delta_params affected_by_baseline_ach -> delta_depends_on_baseline_ach better represents that they describe the delta the intervention adds (and whether that delta depends on baseline ACH).

…ention parameters

… it with intervention documentation and Roxygen

… riskiness by setting median

cwhittaker1000 · 2026-05-29T21:10:29Z

+    volume_per_person_workplace = 27, #10m^2 *2.7m
+    volume_per_person_school = 10, # 3.33m^2 *3m
+    volume_per_person_leisure = 8, # 2m^2*4M
+    volume_per_person_household = 50, # 20m^2 * 2.5m


I think this isn't currently connected to the square footage per person assumed for coverage allocation etc (or at least, it matches only through the defaults, a user could change this and it might not change the square footage used in coverage allocation). Need to make sure they're joined up if they're not already!

Geetha: maybe just do square footage per person and multiply by heights.

Charlie: yes I like that idea!

…xploring structures of various interventions

making branch appear online

588ab04

geethaj1 added 2 commits February 10, 2026 23:27

exploring normal and lognormal ACH distributions and the effect of th…

299da28

…ose on the riskiness distribution

Implementing replacement of directly inputting riskiness with ACH -->…

691b978

… riskiness.

geethaj1 and others added 4 commits February 11, 2026 14:25

updates to replace efficacy inputs with ach inputs

e031316

moving from /man to /R

bbb5b9a

Update setting_specific_ach.R

65eb2ce

Update setting_specific_ach.R

314bf02

geethaj1 reviewed Feb 12, 2026

View reviewed changes

Comment thread R/parameters.R Outdated

geethaj1 added 2 commits February 17, 2026 13:29

setting up ach to efficacy pipeline with constant and sigmoid options…

3bfcb0f

… for relationship

update parameters for ach efficacy

e11ebbc

cwhittaker1000 commented Feb 17, 2026

View reviewed changes

Comment thread R/parameters.R Outdated

Comment thread R/setting_specific_ach.R Outdated

Comment thread R/setting_specific_ach.R Outdated

Comment thread R/setting_specific_ach.R Outdated

geethaj1 added 8 commits February 18, 2026 09:12

removing duplicate

3297541

adjusting room volume to be calculated using density * number of people

3590308

continuing to add to ach --> efficacy pipeline

2582144

cleaning up errors in code (forgotten closing brackets, parameters fo…

597b108

…r each type of location, etc.)

added based on claude suggestion

aef96ec

Test run with a signmoid function for workplace efficacy + edits to s…

f714a04

…etting_specific_ach script

removing duplicate calculate_efficacy_from_ach call

fac2c6e

moving test scripts to inst

1b0354a

geethaj1 added 8 commits February 25, 2026 16:50

multiple settings

af4b4b0

updates to room volume based on ASHRAE area densities and room height…

5ca5e03

… estimates

minor edit to constant ach script

0d2d6e6

tests with UVC with new ACH pipeline + adjustment to set_setting_spec…

ebcb680

…ific function based on tests

added riskiness plots per location type

4aa74f0

more tests

bedb90c

moved to inst

91b8e28

gathering metrics from tests

6f3ef3e

geethaj1 added 8 commits May 6, 2026 09:11

comment out old code

d58a50c

testing new implementation

bfcb842

edit

23b5956

fixing piping for new intervention implementation

b5e3c25

Moving ach intervention pipeline to interventions.R + run preliminary…

6b3d22a

… tests to check that it works

Commenting out old code, and moving all functions related to interven…

fca4247

…tions to interventions.R

edits from review

621164c

edits based on code review

9c6d950

geethaj1 changed the title ~~updating to directly input ACH and infer riskiness and intervention efficacy from that~~ updating to directly input ACH and infer riskiness. User defined efficacy from delta ACH May 18, 2026

geethaj1 added 2 commits May 18, 2026 12:34

fixing coverage error in code

66fd281

tests

9405742

Apply suggestion from @github-actions[bot]

1adce15

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>

cwhittaker1000 commented May 19, 2026

View reviewed changes

geethaj1 added 8 commits May 19, 2026 12:49

Add scripts for branch-vs-main comparison and intervention testing

eb4f886

deleting old set_uvc code that has been commented out

47c4a66

removing commented out code

fc354b8

documentation added for W-R parameters, volume per person, and interv…

856da6a

…ention parameters

removing all old uvc functions/parameters/documentation and replacing…

7f55fea

… it with intervention documentation and Roxygen

Removing unused wells_riley_reference_ach parameter; always normalise…

5aaaa11

… riskiness by setting median

edits based on charlie feedback, gets user to pass list directly

7fa1b59

cwhittaker1000 commented May 29, 2026

View reviewed changes

geethaj1 added 7 commits June 3, 2026 01:10

update to remove ach = 1 default and force user to enter ach value

99ef18d

removing ach = 1 default

7571977

Adding Roxygen documentation

6e2f4d8

updating testthat scripts to test ACH implementation + added script e…

7bd1c82

…xploring structures of various interventions

adding plots to tests

ad26fe8

addiing plots to tests

4dee30b

updating tests

8785664



		# =============================================================================
		# Helper functions for ACH / efficacy / UV-C conversions

Uh oh!

Conversation

cwhittaker1000 commented Jan 22, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

cwhittaker1000 commented Jan 22, 2026

Uh oh!

geethaj1 commented Feb 11, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

geethaj1 commented Feb 25, 2026

Uh oh!

cwhittaker1000 commented May 19, 2026

Uh oh!

cwhittaker1000 left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

geethaj1 May 29, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

cwhittaker1000 commented Jan 22, 2026 •

edited

Loading

geethaj1 commented Feb 11, 2026 •

edited

Loading

geethaj1 May 29, 2026 •

edited

Loading