Credits: Venkat Raman on LinkedIn

1. The p-value is uniformly distributed under the null hypothesis.

2. You can't accept the Null Hypothesis.

3. In nature, Normal Distribution is not the most prevalent.

4. There is a difference between Normal Distribution and Standard Normal Distribution.

5. The earliest use of Logistic Regression was not for classification but regression. Logistic Regression outputs probabilities not just 0 and 1.

6. The confusion matrix is not named because it causes confusion.

7. Non Parametric does not mean No Parameters.

8. Central limit theorem does not kick in at n=30.

9. Confidence Interval will be always narrower than Prediction Interval.

10. Log transformation will make your data near "Normal" only if your original data distribution was Log Normal.

11. F test in Linear Regression is not useful.

12. The standard error is the estimate of SD.

13. Errors != Residuals.

14. Maximizing the Likelihood is equivalent to minimizing KL Divergence.

15. AUROC is all about ranking and has an interesting relation with Mann Whitney U test.

16. PDF != Probability.