Can you show/share your computations of the confidence intervals? Also, less than 95% intervals might be relevant. (Maybe Dm me if you want to discuss?)
Meanwhile, I did a few calculations based on inputting the numbers as eyeballed from your graphs. The very wide CIs do seem plausible, I also calculated wide 95% CIs with both the normal approx and with bootstrapping … so I’m not saying you made an error.
As per our conversation, can you please update the text (and add a ‘crossout’) perhaps to reflect that you have reported the 95% ranges of the observed data and not 95% confidence intervals?
I think it’s important to get and state our statistics right, for promoting epistemics, statistical literacy, and an empirically-driven environment among the EA community.
(I really appreciate the work you have done on this by the way)
Can you show/share your computations of the confidence intervals? Also, less than 95% intervals might be relevant. (Maybe Dm me if you want to discuss?)
Meanwhile, I did a few calculations based on inputting the numbers as eyeballed from your graphs. The very wide CIs do seem plausible, I also calculated wide 95% CIs with both the normal approx and with bootstrapping … so I’m not saying you made an error.
As per our conversation, can you please update the text (and add a ‘crossout’) perhaps to reflect that you have reported the 95% ranges of the observed data and not 95% confidence intervals?
I think it’s important to get and state our statistics right, for promoting epistemics, statistical literacy, and an empirically-driven environment among the EA community.
(I really appreciate the work you have done on this by the way)