However, some ideal conditions have been assumed including independence of markers and of study subjects. In reality markers will be in linkage disequilibrium and the approximation by an effective number of independent tests is heuristic. Similarly, subjects will be related, if distantly. Results from real data may depart from those presented here if proper account is taken of relationships between subjects. In particular, shrinkage estimation is likely to improve power and correlation, as well as mean square error, by analysing all markers simultaneously rather than each one marginally [37].
