The data presented examine statistical relationships between census (area level) and electorate swings.
-
Show this thread
-
The statistics look quite impressive. But they’re simplistic (not necessarily a bad thing) and have the potential to lead to people putting way more value in their certainty than can be interpreted.
1 reply 0 retweets 0 likesShow this thread -
The r2 (r-squared) values indicate how much the models (2 variables) relate to each other account for or explain the full nature of what’s going on (variance). The r-squared values are low (not always bad). This suggests that there’s more to this than these factors alone.
1 reply 0 retweets 0 likesShow this thread -
A better analysis is to examine something like the Australian Election Study and use a more thorough model to control for more variables. We might find these relationships wash away (or are strengthened) when more information is included.
2 replies 0 retweets 3 likesShow this thread -
There’s no certainty in statistics. Making claims using simplistic models about people (individuals) can cause real harm to communities. Beware what you take from these data.
2 replies 0 retweets 3 likesShow this thread -
Replying to @DrDemography
thanks Liz, I have tried to emphasise as much as possible that this is about the characteristics of electorates rather than actual voter decisions... but possibly I should put in more of an explanation about the r values and strength of correlation?
3 replies 0 retweets 2 likes -
Replying to @NickEvershed @DrDemography
It might be good to add a bit about the strength of correlation, to someone without any stats understanding it might look like all the graphs show the same level of association
1 reply 0 retweets 1 like -
-
Replying to @NickEvershed @DrDemography
Might also be cool to include a model with all the values. Looks like you get an R^2 of about 6, and the things that are still significant with all the controls are coal mines, engagement, migrancy, income/home, and youthpic.twitter.com/4q0jLGzIhC
2 replies 0 retweets 1 like -
Oh, I did this with BMA for
@NickEvershed beforehand :) https://cran.r-project.org/web/packages/BMA/BMA.pdf …1 reply 0 retweets 1 like
Cool! I'd be interested to see the results of a more in-depth model, it looks like there are some interesting correlations there 
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.