The point is, @sarahkendzior didn’t do the subgroup comparisons. She posted an image with a table from the report, including %.
-
-
Replying to @gilliatt @sarahkendzior
No, she did the subgroup comparisons in her tweet. The study is fine.
3 replies 0 retweets 7 likes -
Replying to @zeynep @sarahkendzior
Don’t I see that actual number in the table? It’s a whole table of percentages.
1 reply 0 retweets 1 like -
Yes, the percentage is literally in the table.
2 replies 0 retweets 0 likes -
Replying to @sarahkendzior @gilliatt
Yes but one cannot compare tiny groups (Stein vs Clinton/Trump voters who support Duke) because of margin of error.
3 replies 0 retweets 6 likes -
I will see if I can hunt a longer article on this. The issue isn't there are percentages.
1 reply 0 retweets 1 like -
See calculation. LA has 3 million reg voters; sample you cite is 12. ~Margin of error=+-29% https://www.surveymonkey.com/mp/margin-of-error-calculator/ …pic.twitter.com/DliBrDVDtH
1 reply 0 retweets 8 likes -
Now, ordinarily, you wouldn't even try with such a tiny number but let's go with it to see what it means.
1 reply 0 retweets 5 likes -
That number you cite, 2 out of 12 (17%) means -12 to 46%, at generous 95% confidence level. Zero is within margin.
1 reply 0 retweets 5 likes -
So you couldn't say this number is different from zero. Therefore can't compare to anything else.
1 reply 0 retweets 6 likes
Of course, nobody would even bother with margin of error with 2 out of 12 in a sample of 500 polled from 3 million.
-
-
So the problem isn't the study is flawed. Study is just fine. You always have tiny groups in samples.
1 reply 0 retweets 7 likes -
Problem is drawing any conclusion—including their existence when margin of error includes zero—from tiny subgroup.
5 replies 1 retweet 11 likes - Show replies
New conversation -
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.