Improves distribution of dummy data #2326
Open
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Deploying databuilder-docs with
|
| Latest commit: |
dc68b87
|
| Status: | ✅ Deploy successful! |
| Preview URL: | https://a36ced6f.databuilder.pages.dev |
| Branch Preview URL: | https://drmaciver-field-weightings.databuilder.pages.dev |
rebkwok
reviewed
Jan 14, 2025
| Dummy data generation will generate a larger population and then sample from it | ||
| to improve the distribution of patients. This parameter controls how much larger. | ||
| Lower values will be faster to generate, while larger values will get closer to | ||
| the target distribution. |
There was a problem hiding this comment.
I think it'd be helpful to users to give the default and an idea of what a small/large value is. i.e. 1 means no oversampling, 2 means oversampling by up to 2x the specified population size
| Defines a "weight" expression that lets you control the distribution of patients. | ||
| Ideally a patient row will be generated with probability proportionate to its weight, | ||
| although this ideal will be imperfectly realised in practice. The higher the value of | ||
| ``oversample`` the closer this ideal will to being realised. |
There was a problem hiding this comment.
An example as we have for additional_population_constraints would be helpful. Will this always be a case()?
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This improves the distribution of dummy data through the strategy of oversampling: We produce a larger population of patients than requested, and then sample down to a subset of it according to some weighted distribution of values.
This weighting scheme is calculated in two ways:
It's also a good place to insert heuristics we might want to add (e.g. we could add default age and sex distributions, we could choose what proportion of nullable columns should be null, etc). This doesn't do any of that.
It does come with the cost of making dummy data slower. I think this is an OK tradeoff, and will be improved by anything that improves dummy data generation (in particular future work on making constraints satisfied more often)