Potential discrepancy in epidemic sizes #232
Replies: 2 comments
-
I think the problem is that it's fiddlier to derive growth rate from R for SEIR models (equation 3.2 in above paper), and hence project forward with a simple (1+r)^n function. Instead, I had a go at estimating empirical growth rate from the simulation and comparing with the SIR growth rate from above paper, and looks like theory and simulation match OK (noting I approximate the empirical calculation with an exponential). I also put some people in the initial_conditions <- matrix(
c(S = 1 - 2e-6, E = 1e-6, I = 1e-6, R = 0, V = 0),
nrow = 1, byrow = FALSE
)
# .....
recovery_rate <- 1/7
infectiousness_rate <- 1/7
beta = 2*recovery_rate
beta2 = 0.8 * 2 * recovery_rate
time_max = 100
# .....
# convert reproduction number to SEIR growth rate (from Wallinga & Lipsitch)
convert_R_to_r <- function(R,infectiousness_rate=1/7,recovery_rate=1/7){
0.5*(-(infectiousness_rate+recovery_rate)+sqrt((infectiousness_rate+recovery_rate)^2 - 4* infectiousness_rate*recovery_rate*(1-R)))
}
# function to calcualte empirical growth rate
empirical_growth_rate <- function(x){
log(x[100]/x[51])/50
}
growth1_empirical <- empirical_growth_rate(new_infections(data)[["new_infections"]])
growth2_empirical <- empirical_growth_rate(new_infections(data2)[["new_infections"]])
growth1_empirical/growth2_empirical
#> 1.562317
# calculate daily growth rates
growth1 <- convert_R_to_r(2,recovery_rate)
growth2 <- convert_R_to_r(0.8*2,recovery_rate)
growth1/growth2
#> 1.563595 |
Beta Was this translation helpful? Give feedback.
-
Thanks a lot - that's really helpful! Keeping this issue open for future reference. |
Beta Was this translation helpful? Give feedback.
-
Pasted from Slack:
I’m trying to think through epidemic size expectations in model scenarios that differ only in their transmission rate (or R, if you prefer). I would expect that for an identical number of initial infections, the ratio of epidemic sizes at time
t
for two different transmission ratesbeta
andbeta2
would be((1 + beta) / (1 + beta2))^t
. I’m clearly missing something here, as this isn’t borne out in the reprex below. What am I missing?Useful reference: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1766383/
Created on 2024-03-22 with reprex v2.0.2
Beta Was this translation helpful? Give feedback.
All reactions