We're a little crazy, about science!

Day 233: Coronavirus modeling – Part 6

DeathsModel-NEW

Model estimated death rate plotted in blue compared to the actual death rate plotted in purple. The model predicts deaths based on the previous infected datapoint, so it’s only predicting 1 day ahead, which is not very helpful. However, it lets me check that my model is running correctly.

Another day another unfortunate datapoint for my model validation. On one hand it’s good to be able to further validate my model, on the other, it’s heartbreaking to see it coming to pass. My model prediction isn’t pretty and the trend so far has been very similar to the model. Let’s talk about how we validate the model.

Yesterday we saw my models prediction for the future. However, I realized I didn’t mention how we know it is even remotely correct. This is done by withholding data from our model and then comparing the measured data with our recorded data. I used quite a bit of data to train my model (find the hidden variables needed to estimate the dynamics). So I had a limited validation set.

In the end, I wanted to use this model to predict about one week ahead and my logic was that knowing the need for the coming week(s) would be helpful for hospitals to be able to better prepare. Right now I am adding every new day of data to the plot and that way I can compare it to the prediction my model is making. This is how we check that my model is correct and really it’s the only metric that matters.

So that said, I have six days worth of data my model has not seen. In this case I am looking at my death rate prediction vs the actual death rate and my model does a very good job of predicting these states. Let’s take a look, below is the plot of my model (in blue) the data my model has “seen” in purple, and the new data that the model has never seen in orange. As I mentioned, this is six days worth of new data the model has never seen so this is basically hitting the goal that I had with this model. Of course, further validation will be best and even though the assignment is due today, I plan on continuing the project for a bit just to see how close my model predicts the progression of the pandemic.

Measured deaths that the model has seen plotted in purple, the model prediction plotted in blue, and the new data withheld from my model plotted in orange.

Measured deaths that the model has seen plotted in purple, the model prediction plotted in blue, and the new data withheld from my model plotted in orange.

I made a whole video about this along with a powerpoint and a write up for the class, so expect to see at least the video if they like it and share it on the labs website. Until that happens or something changes with my model this will (probably) be the last update on the COVID-19 Model that I’ve made. It’s been a learning experience for sure, but I am really glad that I did this.

But enough about us, what about you?

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.