Why are they impossible? How can you tell at a glance?

Because, as Heathers and Brown's DEBIT preprint outlines, the means and standard deviations for binary variables are connected mathematically by this formula: SD = sqrt {[P*(1 � P)*N] / N � 1}. In the Imgur tables, all six of the highlighted SD's are wrong in the original (on left), and are correct in the correction (on right).

Screwing up binary variables' standard deviations all the time in every article you publish is just not an error that people normally make.

Heathers and Brown's DEBIT preprint:

https://osf.io/pm825/

Why are they impossible? How can you tell at a glance?

The formula for the SD of dummy variables is sqroot of p(1-p). Thus, it can be double-checked where p (i.e., the mean) is provided for a dummy variable.

Thanks. I really cannot relate to making up something like this, when it is SO EASY to just crunch the numbers in Stata or SPSS. Like, what is the point? These were statistically insignificant controls, anyhow, so why bother changing the SDs?

I'm finding everything about this guy's ineptitude/scam kind of baffling. Like, is he doing this in longhand on an abacus? Is he compulsively making up numbers out of thin air? I don't get it.