Few specific points but I hope the solutions appeared in time for those who tangled themselves up with Q3 befiore they attempted the similar part of the assessed assignment. 

The most common omission was 2(iii) --- you change no statistical properties if you 
rotate all the data onto a full set of principal components (i.e. those corresponding to 
strictly positive eigenvalues) whether calculated from covariance or correlation. I had 
tried to avoid the red herring of different scales of measurements by envisaging that only the size variables were used but some ignored this and went off at a tangent about correlation PCA being preferable with a mix of variables. If you throw away some of the PCs (e.g. those with 'small' eigenvlaues) then you will change the numerical results but only by a 'small' amount. Try it and see.

A few are still giving nonsensible numbers of decimal places --- maybe change the 
number of figures printed in S+ to avod being tempted to copy out nine decimal places in a p-value.