I know we have MSS (mean sum of squares), MSR (mean regression sum of squares) and MSE (mean squared error). Is MSR and MSS the same?
deleted. I dont remember this at all
There is no such thing as mean sum. There is mean square and then there is sum of squares.
SST is the total sum of squares (the “variance” of the dependent variable).
MSS = SST / (n-1)
RSS is the total sum of squares of regression (the “variance” of the explained portion of the dependent variable)
MSR = RSS / k
Hence MSS is Not the same as MSR.
I thought I had Anova fully mapped. Where does MSS fit in then?
MSS would be read as “mean sum of squares” but in fact it is just the variance of the dependent variable. It is the sum of squares of the dependent variable devided by (n-1) degrees of freedom (i.e. the variance formula).
Your table is right, “MSS” is not displayed because it is not used for regression analysis. However we know it is just the variance of dep variable.
Sorry anyways for the confusion.
Ok, but to calculate F we use MSR/MSE. In what formulas/circumstances do we use the MSS?
I thought MSS was the entire “mean square” column you have there?
Total Sum Squares (TSS) = SSE + RSE.
You are saying that TSS is the same as MSS?
No. It is claimed that MSS=SST/n-1, but not sure whether it has a meaningful interpretation.
It is the variance of the dependent variable, just that.