Biden v. Trump Election Mega Thread

[dreng3]

Johnson.
A person literal Russian agents called Agent Cob whose own party were so bad at pre-election finance promises they'd publicly said they'd spend the same lot of money 4 times on 4 different things.
A over-eager wall flower who managed to have her office spell her own constituencies name wrong.
Oh and an ex Banker who helped Trump get elected.

And somehow Johnson was still the worst pick of the bunch.

 

[Dwarvenhobble]

If you have familiarity with biomedical literature such as drug trials, you'd know that a ~30% response rate for scientific studies involving human volunteers is not unsual.

I'm not sure what you're talking about with correlation: you seem to be implying it's the same thing as response rate and it really isn't. How significant a correlation is would be analysed with statistics (e.g. Spearman's or Pearson's tests). A 0.3 correlation may or may not be significant according to sample size, variation, etc.

Yes a 30% negative reaction would be significant in medicine.

An only 30% of people responded after being treated would probably however result in the drug being sent back to the lab to be worked on again to improve it to actually be worked on again.

Also it would be an R = 0.3. as significance

No, I'm saying the link you found is obviously not a reasonable way to draw conclusions on the whole nationwide race because that "one area" Clinton was 20 up in was a specific demographic unrepresentative of the country as a whole. As an anaology, the same logic you're employing here would claim that because Clinton was 80 points up on Trump with African Americans she was therefore guaranteed to win the election by a landslide.

And yet again I've pointed out numerous polls had Trump almost certain to lose. It's not an out there idea to suggest once again the polls are wrong. Even the polls saying Trump will win by a landslide (because some are saying that this time). Part of this is the algorithms applied to try and account for bias which are hard to accurately create at the best of times.

 

[Iron]

Yes a 30% negative reaction would be significant in medicine.

An only 30% of people responded after being treated would probably however result in the drug being sent back to the lab to be worked on again to improve it to actually be worked on again.

Also it would be an R = 0.3. as significance



And yet again I've pointed out numerous polls had Trump almost certain to lose. It's not an out there idea to suggest once again the polls are wrong. Even the polls saying Trump will win by a landslide (because some are saying that this time). Part of this is the algorithms applied to try and account for bias which are hard to accurately create at the best of times.

Doesn't matter who wins.
Fun times are ahead, there are contingency plans in place.
The biden leaks failed, I don't know if the streisand effect will work but it seems to have done little. This may as well be fabrications as we can tell.
No October surprise. Only thing left is, as usual, scaring old people to come out and vote during election night, which will work out pretty well for Republicans because of the early voting for Dems.
If Trump loses AFTER election night, we will see history. Plans had been made for this scenario. It will be... glorious.

 

[Dwarvenhobble]

And somehow Johnson was still the worst pick of the bunch.

Well most of the bumbling is an act.
It's also very easy to win and make you opponent seem bad when you don't have an actual position just oppose.

 

[Dwarvenhobble]

Doesn't matter who wins.
Fun times are ahead, there are contingency plans in place.
The biden leaks failed, I don't know if the streisand effect will work but it seems to have done little. This may as well be fabrications as we can tell.
No October surprise. Only thing left is, as usual, scaring old people to come out and vote during election night, which will work out pretty well for Republicans because of the early voting for Dems.
If Trump loses AFTER election night, we will see history. Plans had been made for this scenario. It will be... glorious.

Oh I feel bad for Americans.

If Trump wins it'll be riots and attempts to depose him with everything from the claims of a mob trying to storm the whitehouse to drag him out to trying to use the 25th amendment to pull him out to another impeachment.

Hell a former Obama adviser was claiming the plan if Trump wins is to call for democrat owned states to succeed from the union so that will be fun to watch.

If Trump loses we get to watch Biden just let the states burn rather than tackling the riots or go full authoritarian crackdown while in reality the democrats play Weekend and Bernie's with Biden. That is as long as Trump doesn't dig in and civil war starts to remove Trump before the contested election results are resolved.

A 30% sample size and .3 correlation would be worth reporting for a number of the physical sciences. In fact, that's pretty damn robust for a single study in a lot of physical-scientific fields.

Only as a "More research needed this may be something" thing not as a valid observation a such

 

[MrCalavera]

My dude in my imagination I see people here cutting ties with their family members over wrongthink and it scares me that politics have so much influence over people.

I'm sure you are as concerned about those hypothetical people, as you are about politicians being hurt by bad thoughts.

 

[Iron]

I'm sure you are as concerned about those hypothetical people, as you are about politicians being hurt by bad thoughts.

what

 

[MrCalavera]

what

You were very concerned about someone thinking bad of Trump, weren't you?

 

[Agema]

Yes a 30% negative reaction would be significant in medicine.

You're talking about a response rate: i.e. you ask someone to take part in a study (/poll) and some say yes and some say no. Then you do the study (/poll) on the ones who say yes. This is not the same as what data you get from those respondants.

An only 30% of people responded after being treated would probably however result in the drug being sent back to the lab to be worked on again to improve it to actually be worked on again.

Also it would be an R = 0.3. as significance

The r value does not tell us statistical significance, it tells us the strength of correlation. Significance is denoted by p value.

Whether a treatment would be accepted depends on the difference between people who receive the treatment and those who don't (placebo, normally). 30% getting a benefit is great if the alternative is zero getting a benefit without treatment.

And yet again I've pointed out numerous polls had Trump almost certain to lose. It's not an out there idea to suggest once again the polls are wrong. Even the polls saying Trump will win by a landslide (because some are saying that this time).

And yet you've pointed out yourself recently (or maybe it was someone else), I think fivethirtyeight had Trump at the distinctly plausible ~30% win chance in 2016. How on earth do you get that, and then claim the polls suggested Trump had no chance?

Let's take a fairly normal scientific paper that does a set of experiments with 7 individual experiments (n=7) compared against 7 individual control experiments, and they test it and get a p value of 0.04 (i.e. a 96% chance the difference between the two is not random). So, it's 96% likely to be true, then? No, it isn't. For various teechnical reasons about the way that statistical tests work, given those small sample sizes, the chance that result is "true" is actually likely to be far lower, perhaps as low as ~50%. Seriously, it's true.

As a scientist, this is stuff I know. There is published data I have out there - honestly collected and analysed - that even I as the person who did the work and wrote it up suspect with hindsight and other publications is "wrong". We scientists read papers and they say stuff and we don't necessarily believe them. Because we know it can be wrong. It's firmly established when enough people have done enough experiments. One little n=7 is worth not that much. 5 papers each with n=7s is an n=35, and that's starting to look good.

And so it is with polls. Any individual poll islow reliability, but if you start looking at multiple polls and taking averages, the results shrink.

Part of this is the algorithms applied to try and account for bias which are hard to accurately create at the best of times.

Sure, but they're also pretty good. If you know things like how various demographics vote, and vote compared to one another, it means pollsters can correct for unrepresentative elements in the data sample. If they repeatedly find in elections they repeatedly underestimate a side by a few percent, they can factor that in. Getting these accurate is tricky, but they're usually more accurate than no adjustment at all. Again, check fivethirtyeight. They not only put the polls up there, but they do a lot of other analysis. One of the things they do is analyse the quality of different polling companies. They then give them scores (A to D) to reflect how good their polls appear to be, and they measure the average error in their polls.

Again, if you're going to bring up science again, science does a lot of this sort of thing, too - attempts to make data clearer or to artificially correct for assumed errors. For instance, when I and most of my peers try to pick out very small electrical currents measured across a cell membrane, we leave it up to an algorithm, and then double check it with visual inspection to remove obvious errors. And we have to set the algorithm correctly. And there IS going to be error on every last measurement any of us do. That's just how it is.

 

[Dwarvenhobble]

You're talking about a response rate: i.e. you ask someone to take part in a study (/poll) and some say yes and some say no. Then you do the study (/poll) on the ones who say yes. This is not the same as what data you get from those respondants.



The r value does not tell us statistical significance, it tells us the strength of correlation. Significance is denoted by p value.

And yet R can be enough to deem the result a significant one in some areas

Whether a treatment would be accepted depends on the difference between people who receive the treatment and those who don't (placebo, normally). 30% getting a benefit is great if the alternative is zero getting a benefit without treatment.

Ah but it would entirely depend on the side effects of the medication. Cost vs benefit. That treatment would only be used if no others worked or were available generally it would be sent back to try and further develop or if no other treatments existed split off into one group researching improvements while others get the working one out there.

And yet you've pointed out yourself recently (or maybe it was someone else), I think fivethirtyeight had Trump at the distinctly plausible ~30% win chance in 2016. How on earth do you get that, and then claim the polls suggested Trump had no chance?

Let's take a fairly normal scientific paper that does a set of experiments with 7 individual experiments (n=7) compared against 7 individual control experiments, and they test it and get a p value of 0.04 (i.e. a 96% chance the difference between the two is not random). So, it's 96% likely to be true, then? No, it isn't. For various teechnical reasons about the way that statistical tests work, given those small sample sizes, the chance that result is "true" is actually likely to be far lower, perhaps as low as ~50%. Seriously, it's true.

As a scientist, this is stuff I know. There is published data I have out there - honestly collected and analysed - that even I as the person who did the work and wrote it up suspect with hindsight and other publications is "wrong". We scientists read papers and they say stuff and we don't necessarily believe them. Because we know it can be wrong. It's firmly established when enough people have done enough experiments. One little n=7 is worth not that much. 5 papers each with n=7s is an n=35, and that's starting to look good.

And so it is with polls. Any individual poll islow reliability, but if you start looking at multiple polls and taking averages, the results shrink.

So yes the polls can be wrong and as you said with hindgsight the info could be wrong. That's why I aid not to trust the polls. In part because it gives false hope or false perception of what to expect.

Sure, but they're also pretty good. If you know things like how various demographics vote, and vote compared to one another, it means pollsters can correct for unrepresentative elements in the data sample. If they repeatedly find in elections they repeatedly underestimate a side by a few percent, they can factor that in. Getting these accurate is tricky, but they're usually more accurate than no adjustment at all. Again, check fivethirtyeight. They not only put the polls up there, but they do a lot of other analysis. One of the things they do is analyse the quality of different polling companies. They then give them scores (A to D) to reflect how good their polls appear to be, and they measure the average error in their polls.

So what did it rate the few polls saying Trump would likely win in 2016?

Again, if you're going to bring up science again, science does a lot of this sort of thing, too - attempts to make data clearer or to artificially correct for assumed errors. For instance, when I and most of my peers try to pick out very small electrical currents measured across a cell membrane, we leave it up to an algorithm, and then double check it with visual inspection to remove obvious errors. And we have to set the algorithm correctly. And there IS going to be error on every last measurement any of us do. That's just how it is.

Yes and my argument is trying to measure humanity like many polls do is a task that will bring in far more error even using algorithms.

 

[Shadyside]

My dude in my imagination I see people here cutting ties with their family members over wrongthink and it scares me that politics have so much influence over people.

That's been a thing since the beginning. Lots of friends and families have been split due to ideologies, politics, religions, doctrines, etc.

 

[Agema]

And yet R can be enough to deem the result a significant one in some areas

"Significant" is a word that has to be used with extreme care, because it implies something very specific about the probability something is true in this field. Correlation can mean something which can be worth commenting on, but without knowing specific context I wouldn't care to say.

Ah but it would entirely depend on the side effects of the medication. Cost vs benefit. That treatment would only be used if no others worked or were available generally it would be sent back to try and further develop or if no other treatments existed split off into one group researching improvements while others get the working one out there.

Well, it's not quite like that, but let's not go down that tangent.

So yes the polls can be wrong and as you said with hindgsight the info could be wrong. That's why I aid not to trust the polls. In part because it gives false hope or false perception of what to expect.

Trust in the polls does not have to be a matter of faith, it's about knowing how polls are done and what the results mean, statistics, etc. It's about margins of error, and confidence intervals, etc. They can be learnt and understood, and their accuracy assessed with calculated estimates.

So what did it rate the few polls saying Trump would likely win in 2016?

The polls are on the popular vote. As the candidate who wins the popular vote can lose the election, this is not information that can be gleaned without further analysis. But Trump won via key swing states that backed him by less than 1%. If the polls said Clinton 45 Trump 41 and the final result came back Clinton 45-42, you'd say the polls were very accurate. But that 1% can be the difference between a Clinton victory into a Trump victory.

Yes and my argument is trying to measure humanity like many polls do is a task that will bring in far more error even using algorithms.

It's not that easy. Sample size has a lot to do with reliability. A poll can easily see opinions of 1000 people in a week. A reseacher might spend the same amount of time getting results from 7 biological cells. (Cells are also very variable, with a lot of activity going on that can't be controlled.) But even if humans are less reliable, that's 1000 people compared to 7 cells.

 

[Avnger]

My dude in my imagination I see people here cutting ties with their family members over wrongthink and it scares me that politics have so much influence over people.

The choice to be a hateful bigot doesn't take away others' rights to free association, amigo.

 

[Dwarvenhobble]

The choice to be a hateful bigot doesn't take away others' rights to free association, amigo.

Yeh so people actually are thinking along those lines and doing it.

Man how they'll feel in 10 to 20 years when politics have moved on to something far different to what we see today and they decided to push away and get rid of their family and friends over something ultimately so pointless and [current year] politics.

 

[Silvanus]

Yeh so people actually are thinking along those lines and doing it.

Man how they'll feel in 10 to 20 years when politics have moved on to something far different to what we see today and they decided to push away and get rid of their family and friends over something ultimately so pointless and [current year] politics.

Families have been having schisms over racism, homophobia, and access to healthcare since antiquity. And probably before.

 

[Dwarvenhobble]

Families have been having schisms over racism, homophobia, and access to healthcare since antiquity. And probably before.

Well as there are black Trump supporters, Gay Trump supporters and presumably black gay Trump supporters I'd say now it would seem more like partisan buzzwords to deem those as the reasons for a schism.

As Biden has said he has no intention to go for Bernie's health care plans or improving that much then that' also quite out the window here too.

 

[Shadyside]

 

[Revnak]

Families have been having schisms over racism, homophobia, and access to healthcare since antiquity. And probably before.

The only good debate in Congress was when two congressmen nearly beat one another to death over slavery I stand by this.

 

[Dwarvenhobble]

The only good debate in Congress was when two congressmen nearly beat one another to death over slavery I stand by this.

You mean like this but about Slavery

 

[crimson5pheonix]

Remember the Biden gaff when he said for cops to shoot people in the legs and it was just a moment of senility?
On ABC tonight.