# Identifying discriminant event subsequences among 3 groups

I am using TraMineR to identifying discriminant event subsequences among 3 groups. In the user guide, it does not mention how to interpret p-value when 3 groups are compared. Here is a part of my output:

Subsequence      Support  p.value statistic index  Freq.1   Freq.2   Freq.3
(I)-(V)-(A)-(A)  0.138   0.003    11.388    48     0.126    0.044    0.242


According to this result, this subsequence discriminates significantly the groups. I wonder if it is possible to know among which groups? Does this p-value means that all groups are different from each other in terms of the frequency of this sequence? I appreciate any help and insight.

## 1 Answer

As explained in Ritschard, Bürgin, and Studer (2013), the discrimination is tested by means of a Pearson independence Chi-square test. It is the independence between the variable "Containing the subsequence" and the group variable that is tested. The shown p-value is the p-value of this independence test. In your case, you get a small p-value that leads to reject independence, which in other words means that the subsequence discriminates between the groups.

The result shows the proportion (Freq.1, Freq.2, Freq.3) of sequences of each group that contain the subsequence. Here, you see that there are clear differences between these proportions. The TraMineR package does not test these pairwise differences. However, it should not be hard to do it by retrieving the variable "Containing the subsequence" with the seqeapplysub function.

• Thank you very much for your response. I am confused a little about the interpretation of the p.value. If it is smaller than .05, it indicates that the (I)-(V)-(A)-(A) pattern happens more frequently in one group than the other group, that is this sequence distinguish between groups. Is this correct? Commented Jul 16, 2018 at 7:56
• It means that the frequency with which the pattern happens significantly differs among the groups. Now, to be different, the frequencies must be higher in some groups than in others. With three groups, you can not exclude that two groups share the same highest frequency. Commented Jul 16, 2018 at 9:56
• Thanks a lot! To apply the Pearson independence Chi-square test outside the TiraMineR, I will need to check if the pattern happened (or not) for each individual in each group, then compare the groups, I believe. So, the values will be 1 or 0 for each individual in each group, and Pearson independence Chi-square test will help me see if groups differs. Then, I can repeat this process for each pattern. I appreciate your help! Commented Jul 16, 2018 at 11:12