good news: my estimate is unbiased (I did a linear regression and OLS is BLUE*)
bad news: my estimand is not a causal estimand (in fact I have no idea what it is)
Idk the notion of estimate bias is statistical and has an easy to derive formal definition. I feel that when ppl write about bias they sometimes mean statistical bias & sometimes causal estimate bias, but if they are psychologists, we can safely bet they have almost no idea what they write about 😂
Yeah, it's true it could reference bias in estimation of the statistical parameter but, also in epi, this is decidedly not what is being referenced sadly...
But I suggest we now refer to that type of statistical bias as, "you know, the TRUE association"
I'm an IRT guy, so "bias" is basically me translating "DIF" into applied researcher half the time (so the clinicians who edit the journals I like to publish in will take it seriously and do something about it, i.e., stop using garbage measures with very large and non-ignorable DIF).
Comments
bad news: my estimand is not a causal estimand (in fact I have no idea what it is)
* every economist’s favourite stats fact
The saddest part is these are real researchers getting paid to do research and they actually have no idea they're talking about the causal effect.
What a tangled web we weave.
But I suggest we now refer to that type of statistical bias as, "you know, the TRUE association"