Pg. 344 of the Quant. Readings of the CFA Curriculum Question 18: The 95% Confidence Interval for the regression coefficient for the pre-offer price adjustment is closest to: a) .156 to .714 b) .395 to .475 c) .402 to .468 d) .434 to .436 My question for the forum is would you use a two-tailed or one-tailed T-test? The book indicated a two tailed in the answer, HOWEVER, on the very next question they are using a one-tailed, OPINIONS? New to the forum, thanks for your thoughts.

You should be using a two tailed test because you are just testing for significance (either way - up or down). What are the details on the next question? Welcome to AF.

C.I.'s are almost always two sided unless they are called something else.

It’s not really a “test” though! I’m impressed with JDV’s restraint.

chrismaths Wrote: ------------------------------------------------------- > It’s not really a “test” though! > > I’m impressed with JDV’s restraint. Must be a good day.

sorry, t value in the t-student table. wow, so once someone has actually looked at the question, please post the reason why for Q18 it’s two tailed and Q19 it’s one tailed.

Be nice and you get results. Be snotty to people who are helping you for free, and you will get told to GFY. Post it up yourself.

I figured it out. Even though your testing a null hypothesis that is one sided, you can find the confidence interval with two-sided.

bison_foilist Wrote: ------------------------------------------------------- > Pg. 344 of the Quant. Readings of the CFA > Curriculum > > Question 18: The 95% Confidence Interval for the > regression coefficient for the pre-offer price > adjustment is closest to: > > a) .156 to .714 > b) .395 to .475 > c) .402 to .468 > d) .434 to .436 > > My question for the forum is would you use a > two-tailed or one-tailed T-test? The book > indicated a two tailed in the answer, HOWEVER, on > the very next question they are using a > one-tailed, OPINIONS? > > New to the forum, thanks for your thoughts. This question is still in the curriculum. The OP was asking why confidence intervals use two-tailed tests, but you use one-tailed tests when calculating test statistics for significance testing. i.e. you use a two-tailed distribution (95% ~ 1.96) for calculating the confidence interval: X+|-StandardError*1.96 simple enough but then you use a one-tailed distrubution (95% ~ 1.65) for calculating the test statistic: (X-Xo)/StandardError to compare to 1.65, and ultimately accept or reject the null. Lastly, you reject the null whenever the absolute value of the calculated test statistic is greater than the critical value – I get that – but are we suppoesd to take the absolute value of the critical value, or can it be negative? How to you know when it’s supposed to be negative?

you usually would not get a critical value to be negative but your calc value (your statistic) could be negative.