What is Two-sample T-test? Easy Explanation for Data Science Interviews

[@emma_ding]

Many of you have asked me to share my presentation notes, and now… I have them for you! Download all the PDFs of my Notion pages at www.emmading.com/get-all-my-free-resources. Enjoy!

[@abhisek-chatterjee]

One small thing Emma. The test statistic T follows t(n1+n2-2) i.e. a t-distribution with degrees of freedom n1+n2-2, not t(n1+n2-2,alpha/2). The former is the distribution of T, the latter is the threshold value of it corresponding to significance level alpha for a both-sided test. It's a small thing, but beginners may confuse, that's why I said.

[@alanpdrv]

Hi Emma. Would you ever make a video on doing A/B/C tests? (or A/B/C/n tests). I recently proposed one during a job interview and have been thinking through the implications of it. Maybe you could talk about multi-treatment effects, the multiple comparisons problem, using a bayesian framework to deal with some of the pitfalls, and maybe comment on why some DSs stay very clear of doing these kind of tests whereas others will very happily embrace them?

[@AllieZhao]

Very clear, I like the way using notion to demonstrate your knowledge.

[@emma_ding]

Thanks, Allie! I'm so glad to hear you're finding it helpful.

[@star_7776]

In 2:24 should not be : If t-statistic < t-critical value, reject H_{0} ? Or I am confused

[@jessicagrant6830]

Hey Emma, something is incorrect in either your test statistic or confidence intervals reported as with the test statistic you conclude to reject the null but the confidence interval would concludes that you fail to reject the null because the test statistic falls within the bounds. These should always provide the same conclusion, can you explain what caused the inconsistency? Thanks!

[@austinhoag5130]

What's the intuition behind the degrees of freedom for the unequal variances scenario? That seems like a hard equation to remember during an interview.

[@user-kv3hu6qe6z]

Hey emma , Since you were coding in python , why didn't you use the scipy library "ttest_ind" method to calculate the welch t-test and just set the variance parameter to FALSE . is it that this straight forward approach is not allowed in interviews . I am new to statistics , so genuinely asking because i do no not know . THAK YOU and great videos

[@Philgob]

very nice