Need Help?

Association Scores: Is there a specific threshold that indicates a strong relationship?

Last Updated: Oct 13, 2016
Our association scores scale from -1.00 (no relationship) to 1.00 (the relationship between a concept and itself).

The following guidelines will help you determine the meaning of Topic-Topic associations:
  • A >0.70 usually indicates a definitional relationship (i.e., between awesome/amazing, or screen/display). If two terms are not synonymous or very close in meaning and have a score this high, this is extremely significant.
  • A >0.20 and <0.70 association score is often where the most interesting findings lie in a project. This is where you should focus your attention in any project. These topics are significantly highly related.
  • A >0.0 and <0.20 association score tells that these two topics are slightly or weakly related.
  • A <0.0 association score (or negative association score) indicates that two topics are unrelated to one another in your dataset.

NOTE: Negative association scores do NOT infer negative sentiment.


Here are useful rules for determining the meaning of Topic-Subset associations:
  • Topic-Subset association scores generally fall within a smaller range that is closer to 0.0 than Topic-Topic association scores. They often fall between -0.30 to 0.30. Scores that are much higher or lower than this range are extremely significant.
  • Topic-Subset association scores have less variance from one another compared to Topic-Topic association scores. A 0.05 point differential between two topic-subset association scores is meaningful and interesting. Differences less than 0.05 scores between two Topic-Subset association scores should be evaluated with some caution as they may be influenced by varying sample size (or document volume) of each subset or topic.

More Support

6560c840000970c2e652c21ccafb3167@luminoso.desk-mail.com
http://assets2.desk.com/
false
desk
Loading
seconds ago
a minute ago
minutes ago
an hour ago
hours ago
a day ago
days ago
about
false
Invalid characters found
/customer/en/portal/articles/autocomplete