Zoorob:
https://journals.sagepub.com/doi/full/10.1177/0003122419895254
Their response:
https://journals.sagepub.com/doi/full/10.1177/0003122419895979
My popcorn is ready...
Zoorob:
https://journals.sagepub.com/doi/full/10.1177/0003122419895254
Their response:
https://journals.sagepub.com/doi/full/10.1177/0003122419895979
My popcorn is ready...
Maybe it's luck, but for something like street crime controlling for temperature in a place like Milwaukee in february, march, etc. strikes me as very reasonable, and probably something they should have controlled for to begin with.
Temperature does makes sense in a place like Milwaukee, imo. Do you know why everyone there is an alcoholic? Because it's so cold 7 months out of the year so all there is to do is stay inside, stay warm, and get drunk.
Then when the summer comes you go outside... and get drunk.
"However, if we replicate our original analysis focusing 20, 40, or 60 weeks after Jude’s story, we observe a decline in 911 calls in each case, either in a convex shape (where calls eventually return to expected rates) or a concave pattern (where calls continue to decline over time)."
This is Trump-esque. They are pointing to results which tell directly opposite stories and saying they all support their story of negative effects. The first model shows negative linear effect and positive quadratic. The second shows positive linear and negative quadratic. They conclude that "we observe a decline in 911 calls in each case". What?? This is wild - they're assuming the rest of the discipline is just not paying attention.
"However, if we replicate our original analysis focusing 20, 40, or 60 weeks after Jude�s story, we observe a decline in 911 calls in each case, either in a convex shape (where calls eventually return to expected rates) or a concave pattern (where calls continue to decline over time)."
This is Trump-esque. They are pointing to results which tell directly opposite stories and saying they all support their story of negative effects. The first model shows negative linear effect and positive quadratic. The second shows positive linear and negative quadratic. They conclude that "we observe a decline in 911 calls in each case". What?? This is wild - they're assuming the rest of the discipline is just not paying attention.
they don't know how to interpret quadratic terms.
Yes, but it appears 3897 is blowing this way out of proportion. The shape of the effect can stay pretty much the same depending on the exact linear/quadratic term, even if they switch signs.
they don't know how to interpret quadratic terms.
Wait, so if the effect of police shootings on violence has a negative linear and positive quadratic shape it's "pretty much the same" as if the effect has a positive linear and negative quadratic shape? It is literally the opposite story.
The response is really passive aggressive and defensive in ways that are unfortunate. This line, for instance: "Modeling these data necessitates accounting for the strong seasonality of crime, which is unaddressed in Zoorob’s comment" fails to mention that Demond et al. ALSO left the seasonality of crime "unaddressed." There's no simple "yes, there was clearly and outlier in our paper, and yes, we also missed this (apparently) incredibly important control variable in our paper, too, but we think the finding still holds blah blah blah."
Yes, but it appears 3897 is blowing this way out of proportion. The shape of the effect can stay pretty much the same depending on the exact linear/quadratic term, even if they switch signs.
they don't know how to interpret quadratic terms.
This is such a strange comment. The shapes are reverse images of each other. Especially if you look at the coefficients in their models. The shapes are completely different.
Think many are missing point of adding temp. Original DPK paper is about calls NET of crime—i.e., expected calls given the estimated levels of crime, which would also account for seasonal variation in crime. Z’s reanalysis doesn’t control for crime in all models (which also makes the raw figures in Z somewhat misleading). Adding temp back is another way to try and capture seasonal variation after Z ignores it.
When you control for crime levels, you are controlling for seasonality in crime.
which makes in unclear what including seasonality in crime actually models which probably causes the instability of signs in their models moving from 20, 40 and 60 weeks
They are modeling nothing, literally nothing changed after the police shooting. If they threw in a cubic term I'm sure they'd be quick to interpret that as well. Signs flip from one model to the next and they don't bat an eye. I realize I sound extreme but I'm just so disgusted that three really good sociologists decided to write this response, and that ASR decided to publish it.