Doctor MJ wrote:Here's a quote from Krugman's op-ed on NYT:
http://krugman.blogs.nytimes.com/2014/0 ... blogs&_r=0Unfortunately, Silver seems to have taken the wrong lesson from his election-forecasting success. In that case, he pitted his statistical approach against campaign-narrative pundits, who turned out to know approximately nothing. What he seems to have concluded is that there are no experts anywhere, that a smart data analyst can and should ignore all that.
But not all fields are like that — in fact, even political analysis isn’t like that, if you talk to political scientists instead of political reporters. So, for example, before glancing at some correlation and asserting causation, you really should talk to the researchers.
This is absolutely the issue here.
That is in fact true. I can only say something about the basketball stuff and the climate stuff they posted, and I have to say that there isn't much quality involved.
For basketball:
I just ran a multiple regression on player RAPM data by using possession based and normalized boxscore values, and I found that STL% is not significant at the player level on offense, and found a +0.16 as the coefficient on defense. Meaning, one standard deviation higher STL% gives about 0.16 standard deviation higher defensive RAPM value. So, the standard deviation for STL% is about 0.7 (it is the weighted standard deviation). What worries me about that article is the fact that the method is only vaguely described, but what I sense from that is that Morris is just using a smaller sample (which may be biased, or likely is biased) as well as having a huge effect of the frequency of the respective events in his results. To test the latter hypothesis, I looked at the standard deviation on the team level per game for each entry and found for this season (as well as the found coefficients by Morris in brackets):
Rebounds: 2.1 (1.7)
Assists: 1.6 (2.2)
Steals: 0.8 (9.1)
Blk: 0.7 (6.1)
Turnover: 1 (5.4)
The comparison here should give an answer. If I find no correlation, I have to conclude that my hypothesis is wrong. But if I find a correlation, I have further evidence that the frequency of the events is actually giving here a misleading result. And well, every person should see that the higher the standard deviation is the lower the coefficient becomes. R is -0.88. Or in other words: About 77% of his findings can be explained by the frequency of the events!
So, there might be some inherent value of steals in comparison to other entries (I found that to be significant for the defense, not significant as said before on offense), but his results as well as the standard deviation of each event on a per game basis actually implies: That a steal seems to be more valuable than a rebound, because a steals happens less often than a rebound.
Btw, your assessment of the Rodman stuff he wrote seems also more effected by the pure effort part by Morris rather than the quality of the research itself. When you read the article series you will find a pretty common theme: Morris is actually trying to prove his preconception. He is not seriously questioning his findings at all, he is not making an effort to check for mere coincendences (that Rodman missed more road than home games has an effect, that Rodman missed games when a backup also wasn't avaialable, that Rodman missed games when the opponents were in average slightly weaker), part of the effect he found is simply based on a biased sample he used. And I also suspect that from the remaining 23% for the steals effect article a big part can be explained by a biased sample. And that's where Morris lacks the necessary skill or honesty in order to provide useful research (and where similarities to Berri are seen from my POV).
Regarding climate science:
Pielke Jr lacks the necessary skills in the underlying physics. In order to fully understand climate science it is necessary to understand things like non-linear dynamics (especially fluid dynamics), you need to understand radiation physics; so, it really helps to have a greater understanding of theoretical meteorology in order to understand the influences of additional energy (temperature is just the measurement of the average speed of particles, so just an representation of kinetic energy) in the atmosphere. And then you have the ocean, cryosphere, land mass and biomass as the other components of the climate system. You can't just simply ignore those things based on weak correlation analyses, especially when you are basing that of on biased samples (like Pielke's studies regaring landfall hurricanes, yeah, they don't happen very often, therefore finding a statistical relationship needs a HUGE amount of data which isn't there yet).
I supsect that the most people simply don't appreciate the difference in knowlegde regarding the underlying physics. There is a Dunning-Kruger effect at work, where a lot of people without that knowledge actual believe they can say something about it, because they have experienced different kinds of weather as well as have seen not accurate forecasts based on the NWP models. Way too often I ran into people trying to argue that those kind of things would "prove" that climate models are wrong (e.g.: they can't predict the weather, thus they can't predict the "weather" aka climate 100 years from now or something like that); completely ignoring the underlying fact that weather prediction is actually a initial value problem while climate is actually a bounderary value problem, thus a really important difference which can't be ignored at all.
Doctor MJ wrote:What's also bizarre is that Silver hired a deep community guy in Neil Paine in addition to Morris, but Paine's primarily writing on baseball right now. If that's Paine's passion that's fine, but we're left with a situation where 538's basketball guy probably isn't their best basketball guy. That is frustrating.
Paine said on APBR that he isn't allowed to write about basketball at that point yet, but that this will change in the future. So, he left out whether that is related to his depature on bbr or based on his possible work as a consultant for a NBA team. Having Paine instead of Morris should greatly increase the quality for that topic.