/sci/ - Science & Math » Thread #13889081

105KiB, 768x1067, screenshot.png

View Same Google iqdb SauceNAO

Anonymous Thu 25 Nov 12:37:34 2021 No.13889081 View Reply Original Report

Quoted By: >>13889091 >>13889098 >>13889275

Somebody help me with interpreting these numbers.
How can the Astrazaneca shot have a negative efficiency (-19%) after 121 days?

Anonymous

Anonymous Thu 25 Nov 2021 12:42:37 No.13889091 Report

Quoted By:

>>13889081
That means it's working. (As they actually intended it to.)

Anonymous

Anonymous Thu 25 Nov 2021 12:47:38 No.13889098 Report

Quoted By: >>13889104

>>13889081
That's not effectiveness, it's just the expected value in the likely range, which includes 0.

Anonymous

Anonymous Thu 25 Nov 2021 12:50:03 No.13889104 Report

Quoted By: >>13889249

>>13889098
>That's not effectiveness
>it's just the expected value of the effectiveness
Good illustration of the way """studies""" and """fact checkers""" work.

Anonymous

Anonymous Thu 25 Nov 2021 14:12:27 No.13889249 Report

Quoted By: >>13889255 >>13889270

>>13889104
It's called statistics. Look at the huge range the value lies in. The expected value is basically meaningless when the range is that large.

Anonymous

Anonymous Thu 25 Nov 2021 14:15:49 No.13889255 Report

Quoted By:

>>13889249
Then what is the point of the study? It derives no meaning

Anonymous

Anonymous Thu 25 Nov 2021 14:21:07 No.13889270 Report

Quoted By:

>>13889249
>Look at the huge range the value lies in.
Why isn't it centered at 0?

Anonymous

Anonymous Thu 25 Nov 2021 14:22:43 No.13889275 Report

Quoted By: >>13889281

>>13889081
A negative effectiveness would mean that the vaccine made you more likely to contract the disease. But look at the confidence interval. This just suggests that any protective effect from the vaccine was gone within a few months.

Anonymous

Anonymous Thu 25 Nov 2021 14:25:53 No.13889281 Report

Quoted By: >>13889289 >>13889308

>>13889275
If I take a placebo treatment for COVID, would the expected effectiveness value be negative? You drones are beyond help...

Anonymous

Anonymous Thu 25 Nov 2021 14:28:56 No.13889289 Report

Quoted By:

>>13889281
Come to think of it, it might, if it increases your likelihood of exposing yourself to an infection you think you're protected against.

Anonymous

Anonymous Thu 25 Nov 2021 14:37:25 No.13889308 Report

Quoted By: >>13889325

>>13889281
I'm not really sure what you're getting at here anon. Are you suggesting that, if something has no effect whatsoever, it's effectiveness should be estimated at exactly zero every time?
What do you think confidence intervals and p values are for, exactly?

Anonymous

Anonymous Thu 25 Nov 2021 14:43:04 No.13889325 Report

Quoted By:

>>13889308
>Are you suggesting that, if something has no effect whatsoever, it's effectiveness should be estimated at exactly zero every time?
If something has no effect whatsoever, and your sample of test subjects is adequate, some will get better and some will get worse, without regard to which group they belong to, so the effectiveness range will be roughly centered at zero.

Anonymous

Anonymous Thu 25 Nov 2021 14:45:16 No.13889331 Report

Quoted By: >>13889335 >>13889352

Reading the posts ITT and then realising at least half of you morons probably have a strong opinion on the vaccine despite demonstrating your absolute inability to make sense of even basic data really drives home how shit this board is.

Anonymous

Anonymous Thu 25 Nov 2021 14:46:44 No.13889335 Report

Quoted By: >>13889338

>>13889331
Seethe harder. I'm still not taking it.

Anonymous

Anonymous Thu 25 Nov 2021 14:48:14 No.13889338 Report

Quoted By: >>13889340 >>13889347

>>13889335
Your canned response is NPC tier. And I don't care whether you're vaccinated or not.

Anonymous

View Same Google iqdb SauceNAO 23524.png, 118KiB, 800x789

Anonymous Thu 25 Nov 2021 14:48:50 No.13889340 Report

Quoted By:

>>13889338
>And by the way, I don't care!!

Anonymous

Anonymous Thu 25 Nov 2021 14:51:10 No.13889347 Report

Quoted By:

>>13889338
>And I don't care whether you're vaccinated or not.
>But I do support cutting off your food supply if you don't do as the government commands.

Anonymous

Anonymous Thu 25 Nov 2021 14:53:25 No.13889352 Report

Quoted By:

>>13889331
... so instead of showing how it is done, i'll write a just write a shitty smug comment, that'll convince them

do you also have a totally scientifically convincing, but secret explanation of why they are getting rid of control groups at all levels? lmao

Anonymous

View Same Google iqdb SauceNAO 3243.jpg, 66KiB, 820x800

Anonymous Thu 25 Nov 2021 14:54:57 No.13889356 Report

Quoted By: >>13889394

I flip a fair coin 800,000 times. I assign 1 to tails, -1 to heads. What are the odds that the measured average value deviates from zero by at least 0.19?

Anonymous

Anonymous Thu 25 Nov 2021 15:10:03 No.13889394 Report

Quoted By: >>13889419

>>13889356
Very small. Do your own hookworm

Anonymous

Anonymous Thu 25 Nov 2021 15:25:49 No.13889419 Report

Quoted By: >>13889454

>>13889394
What's the difference between treating 800,000 random people with placebo, and flipping a fair coin?

Anonymous

Anonymous Thu 25 Nov 2021 15:37:42 No.13889454 Report

Quoted By: >>13889482

>>13889419
You're more likely to get a deviation from 0 with your 800k people because they're all living a different life instead of being an inanimate coin.

Anonymous

Anonymous Thu 25 Nov 2021 15:47:17 No.13889482 Report

Quoted By: >>13889506

>>13889454
>they're all living a different life
I take 400,000 samples from a random distribution and put them into one group. I take another 400,000 samples from the same random distribution and put them into another group. What are the odds that the expected values of the two groups will differ by at least 0.19?

Anonymous

Anonymous Thu 25 Nov 2021 15:57:36 No.13889506 Report

Quoted By: >>13889527

>>13889482
This question is not really about whether one group will outperform the other group or not. It is, rather, how the distributional probabilities of the two groups will differ. This has the effect of making the likelihood of two groups to be a linear function. In the case of an experiment that is a multi-viewed model (say, a classification experiment), the probability of a given classifying group is obtained by a linear combination of its predicted probability and the probability of the observed group being a subset of its predicted probability (the set). In other words, the probability of a given classifying group is determined by the distributional probability of the class.

Anonymous

Anonymous Thu 25 Nov 2021 16:06:45 No.13889527 Report

Quoted By: >>13889548

>>13889506
>This question is not really about whether one group will outperform the other group or not
Yes, it is. If you're testing the efficacy of a medication, the question is whether the medicated group will have better outcomes than the unmedicated group. Now suppose my medication is effectively placebo. I take 400,000 samples from a random distribution and put them into one group. I take another 400,000 samples from the same random distribution and put them into another group. What are the odds that the expected values of the two groups will differ by at least 0.19?

Anonymous

Anonymous Thu 25 Nov 2021 16:19:23 No.13889548 Report

Quoted By: >>13889559 >>13889568

>>13889527
Depends. Would you expect the answer to be the same whether 397,000 people in group A were infected compared with only 2 people in group A being infected?

Anonymous

Anonymous Thu 25 Nov 2021 16:24:22 No.13889559 Report

Quoted By: >>13889573

>>13889548
>Would you expect the answer to be the same whether 397,000 people in group A were infected compared with only 2 people in group A being infected?
No, but we're trying to determine the odds that by sheer chance, I just so happened to pick so many more infection-prone people for one group than the other, that there's a large measurable difference in outcome even with a sample of N=1,684,958.

Anonymous

Anonymous Thu 25 Nov 2021 16:27:52 No.13889568 Report

Quoted By: >>13889573

>>13889548
Incidentally, the actual groups being compared in the relevant part of the study included 53,060 people. The number of infections were 86 and 26, respectively.
Funnily enough, adjusted for age and date, the estimated effectiveness was exactly 0 (with a 95% confidence interval ranging from -55 to 36). The -19% comes after also adjusting for sex, comorbidities and some economic factors.

Anonymous

Anonymous Thu 25 Nov 2021 16:28:54 No.13889573 Report

Quoted By: >>13889575

>>13889559
Wrong N. See>>13889568

Anonymous

Anonymous Thu 25 Nov 2021 16:31:17 No.13889575 Report

Quoted By:

>>13889573
>The -19% comes after also adjusting for sex, comorbidities and some economic factors.
That's called correcting biases in the sample, and reduces the problem to the one I pose, but I accept your correction wrt. N=53,060.

Capcode	All Only User Posts Only Moderator Posts Only Admin Posts Only Developer Posts
Show Posts	All Only With Images Only Without Images
Deleted Posts	All Only Deleted Posts Only Non-Deleted Posts
Ghost Posts	All Only Ghost Posts Only Non-Ghost Posts
Post Type	All Only Sticky Threads Only Opening Posts Only Reply Posts
Results	All Grouped By Threads
Order	Latest Posts First Oldest Posts First

Your latest searches