next up previous
Next: Discussion Up: Predicting the Grouping of Previous: Local Information Predictions

Results

The two models were fit to the data from Experiment 1 and Experiment 2 using the structural equation modeling procedure in SAS (PROC CALIS). The results of fitting the Run--Gap Model and the Local Information Model are presented in Table 1 and Table 2. The estimated parameters and statistics are presented side by side for the two models.

 
Table 1: Comparison of prediction model parameters and for models fit to the data from Experiment 1. 

In Table 1 the Run--Gap model has a fit statistic of 2623 with 4 degrees of freedom. Although this might seem large, recall that the effective sample size is 174,781 separate stimulus response pairs which contribute to a null model of 923,649 with 10 degrees of freedom. The Run--Gap model certainly fits much better than the null model. However, notice that the Local Information (Entropy) Model has a of only 79 with 6 degrees of freedom. Although these two models aren't nested and so cannot be compared precisely in terms of goodness of fit, the difference in is so great that it overwhelms the possible loss of accuracy due to the non-nested nature of the comparison.

Table 2 shows the results of the analysis on Experiment 2. The Run--Gap model has a fit statistic of 388 with 4 degrees of freedom compared to a null model of 541,596 with 10 degrees of freedom. The Local Information (Entropy) Model has a of 59 with 6 degrees of freedom. Again these two models aren't nested, but the difference in between the two models is large enough that the Local Information Model is preferred.

 
Table 2: Comparison of prediction model parameters and for models fit to the data from Experiment 2. 



next up previous
Next: Discussion Up: Predicting the Grouping of Previous: Local Information Predictions



Steven M. Boker
Sun May 14 16:19:40 EDT 1995