Hi, I've done my chi-square test a little different to the IFoA's sol. So I did the same in terms of separating into 10 equal bands and worked out the increment to calculate the ranges eg -17.2% to -13.5%, -13.5% to -9.8% etc. When calculating the expected frequency I use the following formula for say the expected frequency from -17.2% to -13.5%: =[(NORM.DIST(-13.5%, 0%, 2.4%, TRUE))-(NORM.DIST(-17.2%, 0%, 2.4%, TRUE))]*2135. Basically the P(-17.2%<X<-13.5%)*total number. I continue this for the rest of the bands. My numbers each time are v similar to the solution just off eg 340.10 vs 341.85 or 1263.35 vs 1262.64 etc. Finally I sum (A-E)^2/E for all bands and check this against =CHIINV(95%, degrees of freedom =band nr-1). Does my approach seem ok? Thanks, Darragh
Hi Darragh Your approach seems reasonable to me and would have been fine in the exam. Worth noting here that you didn't have to do a chi-square test, producing a graph of the returns or comparing mean/standard deviation of the data and the proposed distribution would have scored the same marks. Your method was fine (and in other questions you may have to do a chi-square test so definitely worth being really familiar with these), just if you found it tricky other approaches were possible here Have a look at the marking schedule in the Examiners' Report for more details. Thanks Sarah
