Scientists from around the world for several months, trying to predict the further spread of the coronavirus using the techniques of statistics and extrapolation of current data about the growing number of cases in the near future. Interviewees RIA Novosti experts believe that to accurately predict the course of events is impossible, but you can improve the predictions by adjusting the mathematical model, specifying the input data and comparing the forecasts with each other.
Mac Hyman, Professor of mathematics at the University Tulainkom in Louisiana, together with colleagues from Georgia state University built its own model of forecasting the spread of the disease in the United States. According to him, accurate forecasting very difficult, even in the case of these known and recurring every year epidemics, like the flu.
"I’d trust the forecasts of the specialists and the models that worked with the Centers for control and prevention (CDC) on the prediction of the development of flu," said Hyman RIA Novosti.
He recalled that the last six years, CDC organizes every year a competition among scientists for prediction of how will the annual influenza epidemic.
"the Community of scientists engaged in modeling and was surprised at how difficult it was to accurately predict even recurrent infection transmitted by airborne droplets. It took several years before a reliable model, confirmed in the last few epidemics," said Hyman.
Every year since 2014, more than 20 teams did weekly forecasts, shared techniques and worked on improving the simulation. They met in person every summer, shared ideas and jointly analyzed the arguments "for" and "against" the different approaches. The simulation results have been published in reputable scientific journals.
"last year funding was provided to five teams that have developed a statistical program for physicians. This year the team switched to COVID-19," said Hyman.
He said that in 2019, the team at Los Alamos laboratory received first and second place in the contest according to the prediction of influenza epidemics.
"If I had to choose only one model, it would be my first choice," said Hyman.
According to Hyman, it is important to clearly understand what the assumptions underlying the method of calculation.
"I wouldn’t trust any of the models, where the methodology and assumptions of which are the forecasts that have not been explained clearly," he concluded.
Gerardo of Chowell Puente, Professor of epidemiology and biostatistics at the University of Georgia and head of public health Department of the same University, believes that already accumulated some knowledge that can be�� was to understand the performance of the model in hindsight.
"I would choose a model that to date has given the best predictions," said Chowell Puente RIA Novosti.
From his point of view, developed by his team’s own model "quite well" has shown itself in the short and medium-term forecasts.
"I wouldn’t trust a single model: a useful comparison of different models", – said the Agency Graham Medley, Director of the Center for mathematical modeling of infectious diseases at the London school of hygiene and tropical medicine.
"All models make different assumptions, and we do not know which assumptions are correct. Therefore, it is best to use all models for comparison. Using numerous models, we can compare the results on specific issues. We can also see how dramatically different model," said Medley, who advises the British government on the proliferation of COVID-19.
"If all the models agree, then we can be sure of their accuracy. If they differ, we can begin to understand what assumptions are important," he explains.
Max Lau, Professor of biostatistics and bioinformatics at Emory University in the United States, believes that in each model, you can find favor.
"can’t tell which model is the best. I think they are all useful in different scenarios," – he told RIA Novosti.
According to Lau, to predict based on the model of how to lift the quarantine, we must note that initial work on modeling tend to focus on hypothetical scenarios. For this purpose, the parameters that you entered "manually".
"it is Important to conduct statistical analysis of model parameters using data on the outbreak (of the disease – ed.) to better understand "real" scenarios. That is important-driven statistical model" says Lau.
In addition, according to him, for lifting the quarantine is important to consider how different categories of the population are in contact with each other.
"Opening the economy, it is important to take into account (in model – ed.), trends in social contacts in the community – for example, how to communicate with each other in different age groups of population", – concluded Lau.
