Kepler
Si certus es dubita
Re: 2020 Democratic Challengers V: Bye Felicia
May I point out that being 1% different when you are at the ceiling of a scale means you are all being squeezed together. There is literally nowhere to go past 100% so you just don't have any "room." 1% difference when you are each in the top 5% means you are differing by TWENTY PERCENT of the available space. That is a huge difference. The scale is attenuated.
But beyond that I think the scaling of the test made no sense. It would have been far better to have a completely different scoring model which allowed an open-ended result. Then correspondence to candidate would have been allowed to float freely rather than be restricted to a highly theoretically problematic and artificial [0 ... 1] range.
And of course that's not even mentioning that alignment politically is multi-dimensional and doesn't work well being projected onto a simple number line.
May I point out that being 1% different when you are at the ceiling of a scale means you are all being squeezed together. There is literally nowhere to go past 100% so you just don't have any "room." 1% difference when you are each in the top 5% means you are differing by TWENTY PERCENT of the available space. That is a huge difference. The scale is attenuated.
But beyond that I think the scaling of the test made no sense. It would have been far better to have a completely different scoring model which allowed an open-ended result. Then correspondence to candidate would have been allowed to float freely rather than be restricted to a highly theoretically problematic and artificial [0 ... 1] range.
And of course that's not even mentioning that alignment politically is multi-dimensional and doesn't work well being projected onto a simple number line.