Timeline for Why does R t.test()'s confidence interval shift slightly w.r.t. alternative hypothesis?
Current License: CC BY-SA 4.0
11 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Oct 28, 2019 at 17:59 | vote | accept | profPlum | ||
| Oct 28, 2019 at 17:59 | comment | added | profPlum | works! cool I still don't understand why one-sided t.test uses norm quantiles but whatever I'll give it to you because you answered most of my question | |
| Oct 23, 2019 at 0:30 | comment | added | Dave | @profPlum What happens if you run my code? | |
| Oct 22, 2019 at 21:03 | comment | added | profPlum | I should mention however that I never get this narrower interval all in one piece, since they are one-sided tests one side is always infinity. If you look at the comments you'll see what I get exactly. | |
| Oct 22, 2019 at 20:59 | comment | added | profPlum | I did and I just reran it to double check, the verbatim R code is in the first code-box of my post (not counting the inline one). | |
| Oct 22, 2019 at 20:37 | comment | added | Dave | @profPlum Please post your verbatim R code. I do not get the confidence interval that you get. | |
| Oct 22, 2019 at 20:37 | history | edited | Dave | CC BY-SA 4.0 |
added 471 characters in body
|
| Oct 22, 2019 at 20:12 | comment | added | profPlum | Thanks that is helpful to know. But no those are the numbers that I get for one-sided intervals (please recheck my post)! That's my point; I don't understand why I get those for one-sided intervals (apparently using norm-quantile) but the wider one (using t-quantile) for two-sided. And to reiterate these are both from the t.test() function. | |
| Oct 22, 2019 at 19:29 | comment | added | Dave | You use the t-test to account for the fact that you're estimating the variance. For the one-sided intervals, I am not getting the numbers that you are getting. My confidence interval is narrower: $( 28.69617 ,38.50383 )$. | |
| Oct 22, 2019 at 18:47 | comment | added | profPlum | So then is only the t-quantile CI correct? Also could you please address why t.test's finite one-sided CI boundaries appear to also use norm-quantiles instead of t-quantiles? | |
| Oct 22, 2019 at 18:24 | history | answered | Dave | CC BY-SA 4.0 |