但我唔想好似某 sdv乜鳩咁賣弄學識,我傾向講吓啲大家本身知知地嘅嘢,但實際 implementation 會有出入。例如普通 linear regression 係用 QR factorization, Boxplot 有得做 skewness adjust 而結果會比普通 boxplot 更可靠, 例如 Anova 係有唔同 type 嘅 sum of squares。呢啲 standard 入門 textbook 同 lecture 係唔會有,但 data scientist 知道會有益嘅嘢