Respected research community.
Hope you are doing fine and safe.
Currently I am working on modeling covid19.
In fact, there are many models (SIR, SEIR, SIRD, etc). I would like to know which one is most suitable to model covid19 and why.
Also I would like to know up-to-where modeling has arrived as many pre-prints pretend estimating/forecasting of both recovery (gamma) and infectious (beta) rates. I spent two months digging on this topic and red many articles.
Researchers are using, generally, Matlab built-in functions for prediction ( lsqcurvefit, nlfit, etc).
My second step is how can we analyse stability to the developed model.
If this is a dynamic system, How can we analyse performance (transient/steady state repsponse) and frequency analysis (Root locus, bode plot, etc).
Thank you in advance.
Ayman
Dear Ayman Al-Khazraji , there are several papers that have been published, which try to forecast cases of COVID-19. I think that there is not the best approach, once several exogenous factors can affect the COVID-19 spread. Nowadays, approaches based on artificial intelligence are being used successfully in the evaluation of COVID-19.
Regarding the stability of the developed model, you should use some performance measures, in order to compare the predicted and observed values, such as RRMSE, MAE, sMAPE. Either way, performance measures will depend on your objectives.
I leave the link of my research below
RG: Article Short-term forecasting COVID-19 cumulative confirmed cases: ...
The final paper will be published in https://www.sciencedirect.com/journal/chaos-solitons-and-fractals
Moreover, in the discussion linked in the following, several papers are being listed in order to help the researchers.
https://www.researchgate.net/post/Predictive_models_to_forecast_COVID-19_cases_Artificial_intelligence_and_classical_approaches
With best regards.
Dear Matheus Henrique Dal Molin Ribeiro, thank you for your reply. I will go through your recent paper which seems to be interesting.
I have a paper about using arithmetics rules for the prediction number of patients Covid-19 in West Nusa Tenggara, Indonesia. I try to predict how economic situations in West Nusa Tenggara during the number of Covid-19 patients is increased.
I have cited a few of them in my paper below. Although there are many of them available. Just search through researchgate or google scholar.
Article IoT-Q-Band: A low cost internet of things based wearable ban...
There are at least two classes of models. Those that are epidemiological and those that are dynamical. The former make predictions based on assumed properties of a virus. Good for explaining an epidemic after it happened and these properties are now known, but not good for forecasting during an epidemic. A couple of modeling efforts use current real-time data, e.g., death rates, or hospitalizations. A good one is http://www.healthdata.org/covid/ . However also check out https://syage-covid19-assessment.com/
Thank you all for replying.
Dear Jack A. Syage , your links were interesting.
I have found this document that brings together all the prevalence studies and calculates the IFR (Infection fatality rate) based on it. It is constantly updated. At the time I consulted it, there was an undercount of 1:25 and an IFR of 0.2%. Link: https://docs.google.com/spreadsheets/d/1zC3kW1sMu0sjnT_vP1sh4zL0tF6fIHbA6fcG5RQdqSc/edit#gid=0
This is my link research if you guys want to see my peprint..
https://www.researchgate.net/publication/340966271_Prediction_of_the_economic_situation_in_West_Nusa_Tenggara_because_of_Covid-19_Aritmatic_rules
Dear Stefan Ziemendorff , thank you for your post.
Some papers are pretending estimating recovery rate using confirmed and death cases. However, they don't explain clearly how ?
To my knowledge, there's neither date nor clear metric to check recovery rate performance. Hope that specialized people comment this point.
Thanks in advance.
Dear Ayman Al-Khazraji , that is of course a very important point for your model. I know that the Robert-Koch Institute is using a mathematic model for recoveries as it is not mandatory in Germany to report recoveries and I know that they changed their calculation two days ago, so there are now more estimated recoveries than before. Unfortunately I could not find the original source - maybee you can. Many greetings from Peru!
Preprint Multiparametric computer simulation estimating the hidden va...
I've done both SEIR and forecasting models, my later is an asymmetric Gaussian model. SEIR handles recovery through differential equations. I use deaths to compute cases using values of 2.5 weeks from infection to death and 3.5 weeks from infection to recovery. This is explained in the following:
https://www.medrxiv.org/content/10.1101/2020.05.16.20104430v2
and an improvement to that model in:
Preprint A Statistical and Dynamical Model for Forecasting COVID-19 D...
I also run a blog at
https://syage-covid19-assessment.com/
Perhaps some of these might help. They also summarize recent covid models