Table 3. Significant sign triplets from the log-likelihood ratio (LLR) measure of association for trigrams.
| Sign Triplet | Rank | Frequency | Significant | Rank | LLR Value |
| (Naive) | (EBUDS) | Sign Triplet | (LLR) | ||
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
We use the null hypothesis, 
. The 20 most frequent sign triplets (first column) are compared with the 20 most significant sign triplets (third column). The most frequent sign triplets are not necessarily the most significant sign triplets, as measured by the log-likelihood ratio measure of association.