kdorheim
commented
3 months ago Nicely done! The fit does look pretty reasonable, but there are a few more robust ways to evaluate fit which we haven't covered yet. I use the "observed" data here to refer to the example data set which is a data set that I generated, so I know exactly the correct parameters :)
Some things to look at
- What is the RMSE (root MSE) between your final fitted model & the "observed" data. How does this compare to the "observed" data set's variability?
- Could you plot the residuals (difference between estimated values and fitted values) over time? Is there a trend there?
Because I generated the data set we've got model which is $$y = 3 + 2.5 x + 0.5 x^2$$
How do your fitted values compare with these? Any idea as to why they might differ? Do you have any concerns with the parameter estimates you got?
I used optim to fit the given model to the given data in Issue 12, detailing my work and results in an Rmd file. The example data that the model was fitted to is also included in this PR