This is the last update that I think we'll need to make for the current round of papers. We observed from the simulated policy diffusion cascades that the exponential distribution did not fit very well, and really can't fit a distribution with any bell-like shape, due to its monotonically decreasing density. Since you've implemented profile ML estimation for the exponential, would you next implement lognormal? Note, if y is the vector of diffusion times this is just calculating mean(log(y)) and var(log(y)), since MLE for the log normal is equivalent to normal mle for the log of the variable.
This is the last update that I think we'll need to make for the current round of papers. We observed from the simulated policy diffusion cascades that the exponential distribution did not fit very well, and really can't fit a distribution with any bell-like shape, due to its monotonically decreasing density. Since you've implemented profile ML estimation for the exponential, would you next implement lognormal? Note, if y is the vector of diffusion times this is just calculating mean(log(y)) and var(log(y)), since MLE for the log normal is equivalent to normal mle for the log of the variable.