STA 106 Lecture Notes - Lecture 11: Likelihood Function, Analysis Of Variance, Complement Factor B
Document Summary
If normality of errors, or equal variance by groups are violated, we can transform our data by yij. We have already seen the following transformations. i. e. adding a constant y ij = yij + a: multiplying by a constant y ij = byij. Linear transormations change the variance (larger or smaller spread) and the location (shifts the entire dataset). What we want: if our data (eij) is non-normal, we want a transformation that makes it look linear. We pick the "best" value of a constant for this equation. How to choose : correlation of normal distribution (qq plot). This gives a which has y ij give the. "best" qq plot: p-value for shapiro-wilks (sw). This picks the which gives the largest p-value for. Sw. i. e. if it picks = 1. 5, This uses that yij n ( i, 2. The "likelihood" function is (because yij n )
