tag:blogger.com,1999:blog-6894866515532737257.post4790568570250194105..comments2024-03-28T21:59:14.517-07:00Comments on Probably Overthinking It: Generational changes in support for gun lawsAllen Downeyhttp://www.blogger.com/profile/01633071333405221858noreply@blogger.comBlogger3125tag:blogger.com,1999:blog-6894866515532737257.post-14547124094209790722018-04-24T14:12:28.831-07:002018-04-24T14:12:28.831-07:00Thank you for the clarification - I misinterpreted...Thank you for the clarification - I misinterpreted twenties as the age, instead of the decade of birth.Anonymoushttps://www.blogger.com/profile/00027883649876487686noreply@blogger.comtag:blogger.com,1999:blog-6894866515532737257.post-66855028646770841272018-04-24T13:57:13.538-07:002018-04-24T13:57:13.538-07:00The cells you are looking at are a single run, so ...The cells you are looking at are a single run, so they'll be different every time the notebook runs. The coefficient in Out[98] is a log odds ratio. The offset in Out[103] is a difference in percentage points. The median of that difference is -9.8 percentage points, as shown in Out[108].<br />Allen Downeyhttps://www.blogger.com/profile/01633071333405221858noreply@blogger.comtag:blogger.com,1999:blog-6894866515532737257.post-13957453592875009202018-04-24T13:46:58.957-07:002018-04-24T13:46:58.957-07:00I'm having a little trouble connecting the dot...I'm having a little trouble connecting the dots between your comment "people born in the 1990s are 10 points less likely" and the ipynb that shows the coefficient for "twenties" as -0.0236 at Out[98] which eventually resulted in an offset of -0.473421 at Out[103]. I know you re-sampled to get to the confidence intervals, but it also seems odd that the twenties interval was about as wide as the other ages despite having a z score of -0.531 in the original regression and the thinnest base at about 500 respondents.<br /><br />Regardless of my slow understanding of this, I also worry that the GSS is not a good way to measure the political trends of all ages. Yougov or Google search trends may have less bias because they are not administered or done in-person, in private with less social norm pressure. Anonymoushttps://www.blogger.com/profile/00027883649876487686noreply@blogger.com