Maximum likelihood and experimental data
- To: mathgroup at smc.vnet.net
- Subject: [mg114804] Maximum likelihood and experimental data
- From: max spoor <max.spoor73 at gmail.com>
- Date: Fri, 17 Dec 2010 03:31:22 -0500 (EST)
Dear all, I am new to Mathematica and have a basic question about comparing the fit of different models to experimental data on actual frequencies of play in 2x2 games (e.g. I compare predictions like [0.439; 0.604] to frequencies like [.455; . 636]). Some of the models are non parametric, while others are, so I would like to "punish" the latter for abundance of parameters, say using the Akaike information criterion for likelihood estimation. My question is what distribution should I assume about the underlying data? Also, is there a requirement in terms of the minimum number of observations? Any suggestions or code example will help! max