[ad_1]
INFECTIVE NUMBER INCERTED: In recent days, FHI has released new figures on the spread of infection in Norway. They may seem safe, but there is actually a lot of uncertainty surrounding the current situation. Photo: Gisle Oddstad / VG
The so-called reproduction figure shows how many people with a crown are infected. But what does it really mean when the number rises or falls marginally?
Very little, at least if you are looking to know something about the infection situation. Today, statistical experts VG has spoken to.
The fact that the figure fell below 0.7 has been the main argument that the government has partially reopened Norway after closing the partnership for the first time.
Since then, daily FHI reports have shown that the R number has fluctuated up and down between 0.49 and 0.7. In parallel, Norwegian media, including VG, have stated on their covers that the R number is increasing or that the R number is decreasing.
Data on the R number:
But the small changes that appear in the daily reports of FHI are not the change in the infection situation from day to day. Rather, it is a daily correction of the infection rate for the entire period from April 20 to May 8.
– The figure is suitable for evaluating the usefulness of the measurements, as well as for evaluating whether new ones should be inserted. But it’s bad to say something about how many infected people are still transmitting today, NTNU computer science professor Dag Svanæs tells VG.
– No conclusions can be drawn
Svanæs has previously written a chronicle on VG where he criticized the FHI communication around the century.
– FHI produces an average value for six weeks as if it were the current situation. It is as if the Meteorological Institute said “today’s temperature in Oslo is five degrees” and means the average since March 15, when the outside temperature is 20 degrees, he wrote.
At this time, FHI’s reproduction figures showed an average value for the entire period since March 15, and said even less about the current situation.
Now they have three so-called R numbers to trust. R0 (infected before March 15), R1 (infected from March 15 to April 19) and R2 (infected after April 20). Based on this, they compile the number they use, “R-eff”.
The goal of the figure is to help authorities make decisions to achieve their goal: to avoid tragedies and overloads in the health care system. In other words, a main point of the model is to say something about the future.
This is one of the reasons why FHI has chosen to adapt the model to hospital admissions and not to the number of infected, although the latter would give a figure that says more about the state at this time.
Sweden and Germany use that model. The weakness is that it is vulnerable to changes in testing criteria. Hospital admissions provide a stronger, but at the same time, a delayed figure, which, among other things, aims to provide more secure future forecasts.
FORECAST: FHI uses the R number to tell how the epidemic will unfold three weeks in advance. The graph shows the forecast for the number of hospital admissions as of May 29. Photo: FHI
Great uncertainty
The colored area around the lines on the FHI charts shows confidence interval, which says something about the uncertainty around R.
The less uncertainty, the more data you have. Therefore, FHI is relatively unsure of the reproduction rate between March 15 and April 19, where they have a lot of data, while they have more doubts about the R figure after April 20.
In its latest national report, FHI writes:
“The R2 estimate is still uncertain, as there is little information on hospital data for patients who received the virus on or after April 20. People infected on April 20 and who have to go to the hospital will probably only be admitted on May 2. ”
UNCERTAINTY: FHI also prepares an R number based on the number of infected. They use this, among other things, to verify that the main figure, R-eff, is not entirely on the carry. Uncertainty (confidence interval) increases the closer you get to today’s date. Photo: FHI
0.49 is the FHI R figure presented in their most recent daily report. The figure has a 95% confidence interval (0.0-0.95) and is valid from April 20 to May 8.
Therefore, FHI believes that there is a 95 percent probability that the R figure is somewhere between 0.0 and 0.95. This means that those with relatively high certainty believe that the epidemic has been on a downward curve between April 20 and May 8.
On the other hand, there is considerable uncertainty as to whether the actual R figure is 0.49. Therefore, Svanæs believes that it is the upper limit value that is most interesting to follow for most people.
– As long as this is less than 1, it means that you are almost certain that the epidemic is on a downward curve, he says.
Could it be that R is actually much higher?
Svanæs is not alone in criticizing FHI’s dissemination of the R number.
In early May, two UiT researchers Martin Rypdal and Kristoffer Rypdal published a research note criticizing the use of the FHI model. They thought that the model could not catch up with recent developments and show what the actual reproduction rate is now.
The UiT researchers’ model calculations showed that development in recent weeks could, on the contrary, be interpreted as increasing the spread of infection.
For VG, Rypdal explains that FHI has addressed the main problem with the new R number. However, his criticism of the government’s spread remains unchanged.
“Reports say there is considerable uncertainty about R2, but that is not what they communicate at press conferences every day,” he says.
He believes that those who do the calculations themselves are clear and clear about the uncertainty. Among other things, Arnoldo Frigessi, one of the FHI model developers and professor at the UiO Biostatistics Department, declared on Twitter that the absolute R number “is of very little interest”.
– On the other hand, press conferences give the impression that you know with great precision what R is, since you enter a number of two decimal places. But in reality we are very uncertain. Most of them indicate that R is below 1, but slightly more than we really know very little about, says Rypdal.
– You want to improve
FHI believes that they have been open about uncertainty at all times.
– What does it really mean that the R number decreases or increases in one of your daily reports?
– Daily changes are small adjustments, because the model is being calibrated with another data point. Estimates are uncertain and have large confidence intervals. Therefore, it is not possible to draw conclusions about trends by, for example, a change from 0.55 to 0.49 from day to day, says Birgitte de Blasio, department head for method development and analysis at FHI a VG.
– Do you see that this can be misunderstood by those who do not have a good knowledge of statistics?
– Much attention has been paid to the infection rate. It’s worth noting that our reports not only focus on estimating R, but contain information on plausible development in the number of hospitalized / intensive care patients in the coming time, de Blasio says.
– We explicitly write in reports how the R value should be interpreted. But we always want to be better at communicating our results and considering how to present and publish them, he says.
Trade cooperation: discount codes
[ad_2]
