Shame - SHAME on 538 for covering Rasmussen Reports' deceit.

[speak easy]

Rasmussen Reports has a well known pro GOP sampling bias. Everyone knows that, and 538 compensates for it when added their numbers to aggregates.

https://projects.fivethirtyeight.com/biden-approval-rating/

No amount of compensation however, can excuse results like their numbers for March 3-7 -
Biden approval +0 (49/49)

Rasmussen have a 'unique' method of distributing undecided/don't knows
Their numbers of Biden's disapproval are so out line with reputable pollsters, it beggars credible belief.

But that is not what this OP is about.

Yesterday, I reported posted that Rasmussen's 'polls' had Biden's net approval falling after the stimulus checks started to be received - a result that exposes them to ridicule.

March 10-14 +6 (52/46)
March 15-17 +4 (51/47)

https://www.democraticunderground.com/100215243874

Yes, there's sampling bias, sampling error, and then, at a 95% confidence interval, fraud.

It is not that Rasmussen has a pronounced GOP bias, it is that they appear to operate as a propaganda outfit. As 538 themselves have noted, Rasmussen's results 'magically' fall into line with reputable pollsters on GE eve.

And here we are.

But that is not what is OP is about.

No, today 538 have restated Rasmussen's recent polling

March 11-15 +4 (51/47)
March 16-18 +4 (52/48)

It's not only Rasmussen who are cooking the books.

Rasmussen Reports are a rogue pollster, and 538 are covering for them.

[brooklynite]

1. Why are you blaming 538?

They report polls. They also evaluate and rate the validity of pollsters. You can then evaluate for yourself.

Personally, I'd prefer not to have someone else decide what I should and should not be exposed to.

[speak easy]

2. I am criticizing them for giving credibility to a propoaganda outfit.

Maybe 538 might like to start adding polls by Dick Morris into their aggregates. I suspect he would welcome the invitation.

[uponit7771]

10. ..👍🏼

[Bev54]

3. I agree they should not be apart of any equation in real polls

[PortTack]

4. Agree..other pollsters including the economist does not put razzzmesin in their average

[Indykatie]

5. 538 Does An Adjustment For Every Poll They Record in Their Tracking So They're 2 Sets of Numbers

Everyone knows Rass is shitty and its main goal is driving a narrative for the GOP. Folks treat their results accordingly. I have not heard any Media who is not in the right wing bubble cite Rass's results when discussing Biden's approval. They all generally speak of his approval being "close to 60%" when not referring to a specific quality poll.

[speak easy]

6. You can not adjust for poll biases,

when the bias is unrelated to data. When books are cooked, what matters is the chef, not the recipe.

[Cicada]

7. 538 adjusts for partisan bias so it's wise to include rasmussen

A candidate leading by 80 on rasmussen is more likely to win than someone leading by only 10. That is predictive information so it should not be discarded.

[speak easy]

8. As stated above, you can not adjust for partisan bias

when that bias is independent of the data. When books are cooked, what matters is the chef, not the recipe.

[Cicada]

13. It still has non zero predictive validity

538 adds up a thousand bits of data to get a prediction. Even a measure with only a two percent correlation with the correct outcome has value. I think Rasmussen data has non zero positive correlation with the outcome. Don’t you?

[speak easy]

14. If you arbitrarily play around with the numbers,

within the MoE, 'it still has non zero predictive validity', but fatally compromises a tracking poll.

[WarGamer]

9. I'd take it up with Nate...

Apparently he sees some redeeming characteristic about them.

[speak easy]

11. I don't have to,

it's on DU

[Celerity]

12. If you go to Adults (and not All polls) at the link, it pulls Rasmussen out, and the change is 0.1%

[speak easy]

15. As stated above,

Arbitrarily playing around with the numbers, within the MoE, fatally compromises a tracking poll. When cooking the books, it is the chef that matters, not the recipe. You cannot correct for partisan bias when that bias is unrelated to the data.

[Celerity]

16. I am not going to lose sleep over a tenth of a percent. Bigger fish to fry.

Cheers

[speak easy]

17. If you unconcerned about data/pollster integrity

best school those bigger fish away from polls in general.