TNR is defined as the proportion of correct elements which are correctly retained:
$$
TNR = \frac{\sum_{i=1}^{I}\sum_{k=1}^{K}I(q_{ik}^{t} = q_{ik}^{s} | q_{ik}^{t} \neq q_{ik}^{o})}
{\sum_{i=1}^{I}\sum_{k=1}^{K}I(q_{ik}^{t} \neq q_{ik}^{o})}
$$
where \(q_{ik}^{t}\) denotes the \(k\)th attribute of item \(i\) in the true Q-matrix (Q.true),
\(q_{ik}^{o}\) denotes \(k\)th attribute of item \(i\) in the original Q-matrix(Q.orig),
\(q_{ik}^{s}\) denotes \(k\)th attribute of item \(i\) in the suggested Q-matrix(Q.sug),
and \(I(\cdot)\) is the indicator function.