Comments (3)
- "reporting delays are fixed across age groups and time" Do you mean that the delay "distribution" is fixed?
Yes
- I can't figure out where you're specifying the data when you're fitting the model here. Something missing? Need to be explained a bit more clearly.
If you look at the docs for enw_fit_opts()
you will see it is for setting fitting options and not for actually doing fitting. Data is passed here via the pobs
preprocessed argument. I agree that could be made clearer. An easy change would be to actually write data = pobs
here. Another easy change would be to write some of what I have said here explicitly in the vignette.
- It looks like
data = pobs
argument is missing from themultithread_model
?
This step precompiles the model and so doesn't need data (the actual fitting happens in epinowcast
but combining all he user settings and the applying them to the pre-compiled
object. There is some text around this but obviously it isn't clear enough.
- "To speed up model fitting we make use of posterior information from the previous model (with some inflation) for some parameters. Note that this is not a truly Bayesian approach and in some situations may be problematic." I think this is always problematic, right? You already have some sort of posterior from the data, and you're using the posterior to fit to the same data again (even though you're using the same model). So you're fitting to the same data twice right? I'm not saying you shouldn't do this but the limitation should be made clearer.
The important point here is it isn't the same model. It is a more complex model for which we are using inflated posteriors from a few parameters to speed up fitting (so we get a better idea of what these are). This isn't strictly required but does improve performance a little. The actual use case where it helps more is when doing rolling nowcasting as you can use prior data to inform your current priors. It looks like above could do with a reword.
- Posterior predictions figure is impossible to read... need to zoom in to show a concrete example?
The idea was that people could zoom in on there own time but agree could pull out a sub plot. In general plotting of posteior predictions needs some work. Ideally probably a summary plot and then individual-level plots.
- "As noted using the posterior predictions from the simple model fit above there appears to be a day of the week effect for reported observations" this is really hard to tell...
This should come across by noting the periodicity in the reporting distribution. Not being able to tell this likely comes from above issue using the posterior predictions plot and more hand-holding text being needed. The other thing we really need here is improved visualisation of the raw data (which @FelixGuenther has been working on/off/on and we will hopefully have something for soon).
I think it makes sense to introduce the more complicated model later?
The reason to do this is I think it is a good idea to encourage users to define the model in the order of the process, so generation of reference day counts, reporting distribution by reference date, and then reporting day hazard effects. I agree that the expectation model formula does look a bit scary which isn't ideal. More text would probably help?
I didn't get you were talking about the reference vs reporting day until I got here. I also wonder if it's easier to explain the reporting day effect first and reference day after? I feel like you meant to do this because the reference day of the week model is more complicate
The idea behind leading with the reporting day model is that this effect is more obvious in the raw data and likely has the biggest impact on results (it is also likely much more common - you probably want this in nearly every model). I agree understanding the hazard stuff is more complex though. All modules could obviously do with more description.
U
from epinowcast.
We need to decide if we are aiming to close this out for 0.2.1
or punting to 0.3.0
from epinowcast.
Going to punt to 0.3.0
unless anyone has objections
from epinowcast.
Related Issues (20)
- Update functionality for preprocessed obs
- Expand testing for Renviron as part of persistent compiled Stan models
- Review test warnings and messages HOT 5
- Use epinowcast/actions/install-cmdstan HOT 1
- Double censoring using numerical integration HOT 1
- Figure out where extra options come from in macOS HOT 1
- Touchstone isn't working
- Checks are failing due to the ggplot2 update breaking snapshots HOT 3
- Add better visibility of contributors
- automated workflows that commit don't trigger checks HOT 10
- Linking in docs broken HOT 1
- Pkgdown enwtheme not being used HOT 8
- Incompatibility with scoringutils dev version
- Inconsistent punctuation in function group descriptions HOT 4
- Hex sticker logo has old URL in HOT 1
- Package experience for new user? HOT 1
- Add/Check support for using a `Cmdstan` fit as initial conditions for a new fit + support passing from pathfinder to NUTS
- Map parameter model names to design matrix names + split based on model concepts
- Improve documentation of the formula interface
- Better link model modules with model equation writeup
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from epinowcast.