Accuracy measures in CSS routine

The current CSS routine outputs individual-level accuracy measures. These measure the agreement between an individual's perception of the network (their "slice") with the "true" network.

• • Pearson. This is the pearson correlation between a respondent's slice and the true matrix.

  • Identity. This is a measure of similarity discussed by Zegers and ten Berge (1985).

  • Jaccard. A measure of similarity that ignores 0s. In a cross-tab of the true matrix against a respondent's matrix, Jaccard is a/(a+b+c)

  • Precision. This measures how often, when a person percieves a tie, it is actually there. It is a/(a+b)

  • Recall. This measures how often, when a tie exists, a person perceives it. It is a/(a+c)

References

• Zegers, F. E., & ten Berge, J. M. F. (1985). A family of association coefficients for metric scales. Psychometrika, 50(1), 17-24 [PDF]

• See wikipedia entry about precision and recall measures