Discrete delays have slightly larger log standard deviation than continuous delays

[parksw3]

See scripts/test_discrete.R

• Simulated 1000 delays assuming for growth rates ranging between -0.2 and 0.2. Repeated 20 times per growth rate.
• Directly calculated logsd for continuous and discrete delays

• discrete sdlog is always (?) bigger than continuous sdlog, but the relative ratios are generally small (~2%)
• It's possible that the opposite is true for other distributions? We obviously can't test all possible parameter/distribution combinations but good to know this for understanding fitting results.

[parksw3]

Results for meanlog. But not big enough to worry about.

Something about cases being closer to the either edge depending on the growth rate?

[seabbs]

I think this will be covered in the simulation grid we are planning to include so marking this as for discussion in the paper.

[seabbs]

I think we have covered this (well you have) in the new censoring plot and discussion so closing here.