Skip to content

Commit

Permalink
Merge pull request #76 from The-Strategy-Unit/74_model_naming_guidance
Browse files Browse the repository at this point in the history
added model naming guidance
  • Loading branch information
yiwen-h committed Jul 22, 2024
2 parents 15c6146 + 2e5d657 commit e0f1766
Show file tree
Hide file tree
Showing 2 changed files with 40 additions and 1 deletion.
39 changes: 39 additions & 0 deletions user_guide/naming_scenarios.qmd
Original file line number Diff line number Diff line change
@@ -0,0 +1,39 @@
---
title: "How to name model scenarios"
order: 3
---

# Naming your model runs

Name your model runs carefully. A good name will help you and the model roll-out team understand what's in your model, when it was created and when you ran it.

It's important to realise that at the outputs stage you will have access to the name of the model and the date and time of when it was run, but not the date and time when you created it.

It's possible to select an existing model run from the inputs app and re-run it with tweaked parameters. This could result in model runs with the same name but different parameters, although they'll be differentiated in the outputs app by the date and time they were run.

There are lots of different things to consider when naming model runs and there won't be one approach that suits everybody. In general, we suggest only including things in the model name if they will be meaningful to you in three months' time. For example, the date of creation might be meaningful because you might have designed a two phase process for setting parameters and you want to keep track of which scenario was prepared at each moment. Or it may not be meaningful at all because you are using a different process, like combining the views of several people at a fixed point.

We suggest that you consider including:

* the date of model creation in a simple YYYYMMDD format such as 20240530

* the non-demographic variant in the model, for example NDG2 or NDG1

* simple information such as LOW for a model run that you expect to give a low estimate of demand (e.g. because of high mitigation) and/or V1 for the first model run, V2 for the second, etc.

* hyphens to separate different elements

Example model names might be:

* 20240530-NDG2-LOW (a model with non demographic variant 2 expected to yield a low estimate of demand)

* 20240420-NDG1-V2 (a model with non demographic variant 1, the second model run which has been produced)


## Including words like "final"

Do not use "final" or similar in a model name. The reason for this is because what is a "final" model today may not be a "final" model in the future, and this could cause confusion for you and us. You may wish to include things like "SOC" or "FBC" in the name to indicate where you are in the process, but even then do bear in mind there may end up being two or more "SOC" runs.

## Recording a run as final

In order to make sure that we understand what your final model is, please send the name of the model, plus its run date and time, to your Model Relationship Manager. We recommend that, as with the mitigators, you maintain your own records throughout to avoid confusion.
2 changes: 1 addition & 1 deletion user_guide/setting_the_parameters.qmd
Original file line number Diff line number Diff line change
@@ -1,6 +1,6 @@
---
title: "Setting the model parameters"
order: 1
order: 4
---

The model accepts input parameters, which can be set by users. These are set via the [nhp_inputs app](https://connect.strategyunitwm.nhs.uk/nhp/inputs/). This page lists all the input parameters currently accepted by the model.
Expand Down

0 comments on commit e0f1766

Please sign in to comment.