Skip to main content

Table 4 ANOVA between paired search methods

From: Learning search polices from humans in a partially observable context

 

Greedy vs GMM

Greedy vs coastal

GMM vs coastal

Uniform

3.59e −08 (30)

3.32e −04 (13)

1.90e −01 (2)

#1

5.80e −08 (46)

1.88e −01 (2)

4.58e −06 (28)

#2

3.60e −08 (47)

4.68e −04 (14)

4.54e −06 (28)

#3

3.57e −07 (37)

2.07e −05 (23)

1.25e −01 (2)

#4

6.70e −10 (64)

1.58e −01 (2)

6.34e −13 (107)

  1. Values are expressed as p value (F). The entries in bold are those in which the null hypothesis could not be rejected. The first column gives an indication of the probability that both the greedy and GMM searches are statistically the same (the null hypothesis). This was rejected with a tolerance of below 1%. In the second column, greedy vs coastal searches #1 and #4 are statistically closer than the rest with a p value threshold of 10% required to be able to reject the null hypothesis. In the third column, the uniform and #3 are not statistically different and would require a higher threshold on the p value to be so.