Treating unknown deterministic signals as noise is a handy practice. That's how the GPS system works. Modelling the actual reflectivity of the satellites in order to fully calculate the light pressure terms would be a bear. Easier to just model in process noise. Ditto for modelling changes to the index of refraction in the troposphere.
These unknowns do not look like Gaussian noise if you look at the error residuals. But the system works anyway, albeit not as accurate as theoretically possible. That is, the required error allowances have to be turned up higher than what they would need to be if the noise was Gaussian with the same standard deviation.
I have been scolded by simulation practitioners for not adhering to their principle of using a fixed random seed else it is not possible to compare simulations with different conditions, for example, different numbers of cashiers working the Meijer checkouts.
If you want to understand why simulations are garbage, take a serious course in the theory of differential equations and chaos theory. These models are hyper-sensitive to small perturbations, easily transforming a normally well-behaved system into chaos. Just look at how disastrous Neil Ferguson’s phony pandemic COVID-19 simulations were—textbook examples of how wrong these models can be. Remember this: Worldwide, Ferguson predicted that in the “unmitigated” absence of interventions, COVID-19 could infect 7 billion, resulting in 40 million deaths. Covid-19 was a flu.
I fully agree. Ironically, I used to teach Monte Carlo as a component of financial modelling courses!
I come late to this series. It seems fun, as I've always been interested in methodology. With that in mind, do you happen to have a lecture in your series that discusses alternative philosophies of probability? As an example, Popper's propensity theory (please note I'm not wedded to it - I just mention it because I studied Popper at university). I know this is a bit OT, so no sweat if you haven't! :)
Treating unknown deterministic signals as noise is a handy practice. That's how the GPS system works. Modelling the actual reflectivity of the satellites in order to fully calculate the light pressure terms would be a bear. Easier to just model in process noise. Ditto for modelling changes to the index of refraction in the troposphere.
These unknowns do not look like Gaussian noise if you look at the error residuals. But the system works anyway, albeit not as accurate as theoretically possible. That is, the required error allowances have to be turned up higher than what they would need to be if the noise was Gaussian with the same standard deviation.
I have been scolded by simulation practitioners for not adhering to their principle of using a fixed random seed else it is not possible to compare simulations with different conditions, for example, different numbers of cashiers working the Meijer checkouts.
If you want to understand why simulations are garbage, take a serious course in the theory of differential equations and chaos theory. These models are hyper-sensitive to small perturbations, easily transforming a normally well-behaved system into chaos. Just look at how disastrous Neil Ferguson’s phony pandemic COVID-19 simulations were—textbook examples of how wrong these models can be. Remember this: Worldwide, Ferguson predicted that in the “unmitigated” absence of interventions, COVID-19 could infect 7 billion, resulting in 40 million deaths. Covid-19 was a flu.
William, I thought each red dot represented a new coal fired power plants coming online in China?
I fully agree. Ironically, I used to teach Monte Carlo as a component of financial modelling courses!
I come late to this series. It seems fun, as I've always been interested in methodology. With that in mind, do you happen to have a lecture in your series that discusses alternative philosophies of probability? As an example, Popper's propensity theory (please note I'm not wedded to it - I just mention it because I studied Popper at university). I know this is a bit OT, so no sweat if you haven't! :)
Yep. Class 18, among others. See this list:
https://www.wmbriggs.com/class/
Many thanks!
"alternative philosophies of probability" is the whole course.
Thanks. Got it!