<div dir="ltr">PMI between two random variables can be extended to multiple variables and has some very beautiful mathematical properties.<div style>See the section 2 of this paper:</div><div style><br></div><div style><a href="http://eprints.eemcs.utwente.nl/22600/">http://eprints.eemcs.utwente.nl/22600/</a><br>
</div><div style><br></div><div style><span class="" style="color:rgb(0,0,0);font-family:Arial,sans-serif;font-size:13.020833015441895px"><span class="">Zhu, Zhemin</span> and <span class="">Hiemstra, D.</span> and <span class="">Apers, P.M.G.</span> and <span class="">Wombacher, A.</span></span><span style="color:rgb(0,0,0);font-family:Arial,sans-serif;font-size:13.020833015441895px;background-color:rgb(239,239,255)"> (</span><span class="" style="color:rgb(0,0,0);font-family:Arial,sans-serif;font-size:13.020833015441895px">2012</span><span style="color:rgb(0,0,0);font-family:Arial,sans-serif;font-size:13.020833015441895px;background-color:rgb(239,239,255)">) </span><em style="color:rgb(0,0,0);font-family:Arial,sans-serif;font-size:13.020833015441895px"><span class="">Separate Training for Conditional Random Fields Using Co-occurrence Rate Factorization.</span></em><span style="color:rgb(0,0,0);font-family:Arial,sans-serif;font-size:13.020833015441895px;background-color:rgb(239,239,255)"> </span><span class="" style="color:rgb(0,0,0);font-family:Arial,sans-serif;font-size:13.020833015441895px">Technical Report</span><span style="color:rgb(0,0,0);font-family:Arial,sans-serif;font-size:13.020833015441895px;background-color:rgb(239,239,255)"> </span><span class="" style="color:rgb(0,0,0);font-family:Arial,sans-serif;font-size:13.020833015441895px">TR-CTIT-12-29</span><span style="color:rgb(0,0,0);font-family:Arial,sans-serif;font-size:13.020833015441895px;background-color:rgb(239,239,255)">, </span><span class="" style="color:rgb(0,0,0);font-family:Arial,sans-serif;font-size:13.020833015441895px">Centre for Telematics and Information Technology, University of Twente</span><span style="color:rgb(0,0,0);font-family:Arial,sans-serif;font-size:13.020833015441895px;background-color:rgb(239,239,255)">, </span><span class="" style="color:rgb(0,0,0);font-family:Arial,sans-serif;font-size:13.020833015441895px">Enschede</span><span style="color:rgb(0,0,0);font-family:Arial,sans-serif;font-size:13.020833015441895px;background-color:rgb(239,239,255)">. ISSN </span><span class="" style="color:rgb(0,0,0);font-family:Arial,sans-serif;font-size:13.020833015441895px">1381-3625</span><br>
</div><div style><span class="" style="color:rgb(0,0,0);font-family:Arial,sans-serif;font-size:13.020833015441895px"><br></span></div><div style><span class="" style="color:rgb(0,0,0);font-family:Arial,sans-serif;font-size:13.020833015441895px"><br>
</span></div><div style><span class="" style="color:rgb(0,0,0);font-family:Arial,sans-serif;font-size:13.020833015441895px"><br></span></div><div style><span class="" style="color:rgb(0,0,0);font-family:Arial,sans-serif;font-size:13.020833015441895px"><br style="color:rgb(34,34,34);font-family:arial,sans-serif;font-size:12.571428298950195px">
<br style="color:rgb(34,34,34);font-family:arial,sans-serif;font-size:12.571428298950195px"><span style="color:rgb(34,34,34);font-family:arial,sans-serif;font-size:12.571428298950195px">>Then, is there any solid, mathematically-based association measure between</span><br style="color:rgb(34,34,34);font-family:arial,sans-serif;font-size:12.571428298950195px">
<span style="color:rgb(34,34,34);font-family:arial,sans-serif;font-size:12.571428298950195px">>three or more variables?</span><br style="color:rgb(34,34,34);font-family:arial,sans-serif;font-size:12.571428298950195px">
><br style="color:rgb(34,34,34);font-family:arial,sans-serif;font-size:12.571428298950195px"><span style="color:rgb(34,34,34);font-family:arial,sans-serif;font-size:12.571428298950195px">>Thanks,</span><br style="color:rgb(34,34,34);font-family:arial,sans-serif;font-size:12.571428298950195px">
><br style="color:rgb(34,34,34);font-family:arial,sans-serif;font-size:12.571428298950195px"><span style="color:rgb(34,34,34);font-family:arial,sans-serif;font-size:12.571428298950195px">>Lushan Han</span><br></span></div>
</div>