- y = u + e
- u1 = u2 = ... = uk
在這邊oneway.test就是R做one-way ANOVA的指令,預設會假設每一群的variances都不一樣,這也造成自由度會不太一樣:
- Welch修正,假設每群的varainces不一樣:
- 分子、MSTR (mean square due to the treatment)的自由度=5.000
- 分母、MSE (mean square error)的自由度 = 30.043
- F = MSTR/MSE
- F越小,越能夠拒絕虛無假設
- 假設每群的variances都一樣(var.equal=T):
- 分子、MSTR (mean square due to the treatment)的自由度
K-1 = 6-1 = 5 - 分母、MSE (mean square error)的自由度會是
n-K = 72-6 = 66 - F = MSTR/MSE
- MSTR = SSTR/(K-1)
, where SSTR is the treatment sum of square - MSE = SSE/(n-K)
, where SSE is the error sum of square
沒有留言:
張貼留言