The story that I suspect is trying to be told is that “Republicans cause shutdowns” but that is not what this shows. Using a basic PCC1 shows that these 2 variables aren’t particularly correlated:
With a standard error of +/- 0.302 the signal is technically outside the noise, but it’s not a strong signal and is indicative that different methodologies will give different results.
In other words the data shown does not support the hypothesis. It’s barely correlated and certainly not to enough of a degree to argue causation.
This might show the assumed premise if it included every year without a shutdown. I’m lazy, but I linked the tools so if somebody else wants to show that they can.
This does, arguably, show that shutdown length increases over time independent of controlling parties.
But, as is, it’s a chart of two uncorrelated variables where the author and audience are assuming causation despite available data.
^1. not ideal, especially with such little data, but I’m lazy.^
I agree the story you think the stats are giving are not statistically significant. But I think there is different story… Length of shutdown had drastically increased recently. The problem with stats is the prove truth or show complete trash and both look the same if you don’t have experience.
Length of shutdown had drastically increased recently.
Yes. That is what this is a graph of, but that’s not particularly interesting and doesn’t make for a good graphic. The interesting bit is the addition of uncorrelated variables to imply a correlation/causation that isn’t supported by the data.
The problem with stats is the prove truth or show complete trash and both look the same if you don’t have experience.
Right, which is why I would hope a “data is beautiful” community would be able to recognize trash when it sees it.
The story that I suspect is trying to be told is that “Republicans cause shutdowns” but that is not what this shows. Using a basic PCC1 shows that these 2 variables aren’t particularly correlated:
With a standard error of +/- 0.302 the signal is technically outside the noise, but it’s not a strong signal and is indicative that different methodologies will give different results.
In other words the data shown does not support the hypothesis. It’s barely correlated and certainly not to enough of a degree to argue causation.
This might show the assumed premise if it included every year without a shutdown. I’m lazy, but I linked the tools so if somebody else wants to show that they can.
This does, arguably, show that shutdown length increases over time independent of controlling parties.
But, as is, it’s a chart of two uncorrelated variables where the author and audience are assuming causation despite available data.
^1. not ideal, especially with such little data, but I’m lazy.^
For somebody who knows how to use statistical programs, defining the X and Y values should not be hard…
I agree the story you think the stats are giving are not statistically significant. But I think there is different story… Length of shutdown had drastically increased recently. The problem with stats is the prove truth or show complete trash and both look the same if you don’t have experience.
Yes. That is what this is a graph of, but that’s not particularly interesting and doesn’t make for a good graphic. The interesting bit is the addition of uncorrelated variables to imply a correlation/causation that isn’t supported by the data.
Right, which is why I would hope a “data is beautiful” community would be able to recognize trash when it sees it.