CNN forced to dismiss its own poll after huge majority register positive reaction to sotu

[MindWars]

Title of the thread to which I initially responded (Weatherman2020 created it):

CNN Discredits Own Poll Showing Overwhelming Positive Reaction to SOTU

That thread got merged into this one:

CNN forced to dismiss its own poll after huge majority register positive reaction to sotu

Do you truly not know the difference between "discrediting" a poll and "qualifying" or stating the limit of its scope/applicability?

• Discredit a poll --> To discredit a poll is to show that either the poll itself or a claim it makes is unsound or uncogent. The only way to discredit a poll is to show there are material flaws in its methodology. One can show two general types of failings:
  • Conclusive/inferential failings --> Given the poll's methodology, the claim(s) the poll makes does not follow from the nature and/or extent of the poll's questions or people surveyed.
    • Note: it is very possible, for multi-question polls, that one assertion such polls make be invalid and others they make be valid.
  • Premise failings --> The poll's methodology is invalid in its own right, in which case, it doesn't matter what the poll asserts.
• Qualify a poll --> To qualify a poll is to state, based on the poll's methodology what be the limits of inferences one can validly draw from the poll's findings/results.

While the image CNN depicted of it's poll obviated the absolute need for Chalian and Blitzer to qualify the poll's results/findings, they chose to do so. Were I to speculate on why they did so, I'd say it's because they know the poll methodology -- being an "instant" poll -- does not adjust for imbalances in the political persuasion among the speech's viewer population, thus the poll's sample. It's not surprising that an "instant" poll doesn't thus weight responses so as to adjust for sample composition imbalances with regard to the distribution of political persuasions among the citizenry.

• Chalian: "It's not representative of the country overall it is a poll of people who watched the speech."
• Blitzer: "We're now getting the first results of our instant poll of people who actually watched the President's speech."

​

Take me for example. I didn't watch the speech; I read the transcript of it. Because I didn't watch the speech, I would not be eligible to respond to the poll questions posed to watchers of the speech, and neither would anyone else who didn't watch the speech.

A very reasonable question to ask is: "Is the poll indicative of the citizenry's reaction to Trump's speech?" To put the question differently: "How well do the poll results align with the views of people who "consumed" the speech in some way other than watching it?"

Insofar as the poll only queried people who watched it and not people who read it, there's no way to know the answer to that question because the poll's surveying methodology wasn't developed to answer that question. (Obviously, people who neither read nor watched nor listened to the speech have no valid reaction to the speech that merits being shared; one cannot react to that which one did not experience.) For a poll to answer that question, its developers must:

• Create and execute a random poll. [CNN's poll may have been random with regard to SOTU speech watchers, but it most assuredly was not random with regard to the citizenry who read, heard and/or watched Trump's SOTU speech. It's not because the only people queried were speech watchers.]
• Know the distribution of political persuasions among the citizenry.
• Know the political persuasion of each poll respondent.
• Query enough people for the poll to have a high degree of representational faithfulness.
  
  
  
  
• Weight the poll responses to account for there being more poll respondents of a given political persuasion than is found in the population as a whole.
  • Weighting of responses in the Consumer Survey: Alternative approaches – Effects on variance and tracking performance of the Consumer Confidence Indicator
    
  • Struggles with Survey Weighting and Regression Modeling

Pew provides a general overview of it's methodology here -- Sampling -- and in any given study, it details the specifics of its methodlogy used for that particular study. I don't know how or where CNN (or its polling partners -- CNN itself rarely executes the polls; most of the time, it contracts a polling firm to do it and one term of the contract is that the poll results belong to CNN) discloses its polling methology. (I suggest contacting CNN and simply asking them for it or asking that they point one to it if it's on the Internet.)

From the article:

Chalian was quick to point out that more Republicans were probably watching than Democrats. “So it is definitely an audience that skews a little bit more towards Republicans than Democrats in the Obama years it was the reverse.

I don't know what basis Chalian has for asserting that when a Democratic POTUS delivers as SOTU, more Democrats than Republicans watch the speech. I went to a get together last night. The folks there and whom I know to be Democrats didn't pay much attention to the speech; the Republicans there seemed to periodically "zone in" on parts of it. Based on that, along with the fact that some Democrats announced they'd boycott the speech, Chalian's assertion that the population of speech watches includes materially more Republicans than Democrats seems plausible.

Keep that dream going................