Estimate versus Test Statistic?

acoppock / ri2

Randomization Inference for Randomized Experiments

Other

12 stars 4 forks source link

Estimate versus Test Statistic? #21

Closed jwbowers closed 5 years ago

jwbowers commented 5 years ago

Right now summary(riobject) returns something like

  term estimate two_tailed_p_value
1    Z     0.21                  0

Given the flexibility of ri2, since this is the value of the observed test statistic and is not necessarily a good estimate of any particular target, wouldn't it be better to label this as something like "observed test statistic"?

acoppock commented 5 years ago

I do see your point. in some cases the test stat is the same number as an estimate (as in ate-hat) but other times there no real sense in which it's an estimate of a model parameter, like an f-stat or something.

The main reason to call it "estimate" is to maintain consistency with the broom package -- that way when you call tidy(riobject) you get a column name that matches the broom workflow.

Not sure how to balance these considerations....

jwbowers commented 5 years ago

Good point. Not a big deal. Just mostly an FYI.