Hi all, Just a question on this so im doing September 2021 question 9. I have computed the residual correctly and i want to comment on it. I wanted to look at how many standard deviations away the reisudal was from 0, but im unsure whether this is the standard deviation of y, ie 0.824, or the standard deviation \sqrt (\sigma^2) that i computed in part iii. Please could someone clear this up? Also, what does \sigma^2 actually represent the variance of in these questions, is it \beta? Thanks
Hi Molly I am assuming this is vii b, it is sigma_hat^2=1/(n-2)*(Syy-Sxy^2/Sxx) This is the error variance for the response variable y_hat, your model. The 0.824 is the sample standard deviation. Hope this helps. Thanks Andrea
Thank you andrea, so would i then say that if the residual is greater than 3* sample sd ie 3*0.824 then this suggests non-normality? Thanks
Hi Molly You would use the square root of sigma_hat^2=0.662, ie 0.814 and this would indicate an outlier, indicating non-normality. I was trying to explain that the 0.824, given in the question for y is different, it is the sample standard deviation for the sample values, not for the model, you have to calculate this for the model, using the formula: sigma_hat^2=1/(n-2)*(Syy-Sxy^2/Sxx). Thanks Andrea