Test Families
Picking the right test for the data in front of you: t and z for means, chi-square for counts, ANOVA for several groups, ranks when normality fails. Interviewers ask you to justify the choice.
on this pageshowhide
explore
- T-Tests and Z-Tests5 questions
- Chi-Square Tests5 questions
- Rank-Based Tests6 questions
- Assumption Checks6 questions
- ANOVA and F-Tests5 questions
- Distribution Comparison Tests5 questions
questions
page 2 of 2How would you test whether leading digits in an expense dataset follow Benford's law?
level: seniorimportance: nice to knowfreq 20%
basics
~20 sTally the first significant digit of every amount into nine categories, compute expected counts as N times log10(1 + 1/d), and run a chi-square goodness-of-fit test on 8 degrees of freedom. Rejection flags the data for review, never proves fraud.
Why does Anderson-Darling catch a tail-only difference that a Kolmogorov-Smirnov test misses?
level: seniorimportance: nice to knowfreq 26%
basics
~20 sKolmogorov-Smirnov takes an unweighted maximum CDF gap, and such gaps are naturally tiny in the tails, so its maximum lands near the middle. Anderson-Darling weights the squared gap by one over F(1-F), which explodes at the extremes.
showing 31–32 of 32