1

Is there a help-page for TraMineR:::seqerules? I cannot seem to find it, either in the package nor online. The lack of this help page makes the output somewhat difficult to interpret. For example what do the Conf and Lift columns specify? Below is an example of the output:

Rules                      Support       Conf     Lift
308 (NR)-(QU)-(QU) => (IN)       8 0.61538462 2.666667
153      (IN)-(EX) => (IN)      11 0.55000000 2.383333
394 (NR)-(NR)-(QU) => (IN)       7 0.53846154 2.333333
390 (NR)-(NR) => (NR)-(FA)       7 0.14000000 2.298947
259      (QU)-(EX) => (IN)       9 0.52941176 2.294118
histelheim
  • 4,938
  • 6
  • 33
  • 63

1 Answers1

3

You are right that seqerules is not documented. Help pages are required for public R functions only, and seqerules is currently not a public function of TraMineR. That also is why you need the ::: operator to access it.

The functions returns sequential association rules. The first rule in your example outcome says that when the subsequence (NR)-(QU)-(QU) occurs then it is (generally) followed by the subsequence (IN). This rule has a support of 8 (i.e., it is observed in 8 sequences).

Conf is the confidence, i.e., the probability to observe the conclusion of the rule among the sequences that contain the premise of the rule. For the first rule it is 61.5%.

Lift is the lift, i.e., the ratio of the confidence over the probability to observe the conclusion among all sequences (not only those who satisfy the premise). The higher the lift, the better the rule. A lift less than 1 would means that when the premise occurs, it reduces the chances for the conclusion to occur, and indicates that the rule is of no interest.

Gilbert
  • 3,570
  • 18
  • 28
  • I noticed that there also are p-values for each of the rules - what do these p-values indicate? – histelheim Mar 11 '15 at 00:10
  • I would guess that the p-value tests whether the lift is significantly greater than 1. The lift and the p-value seem to be correlated at a level of roughly -.36 in my data (N=854). – histelheim Mar 11 '15 at 00:42