Where N is total number of pairs, N = 100 in this table.

The 72 pairs that agree do not give us any information on what media is better to detect the bacilli.

For this reason, only we take into account pairs that do not agree, in the analysis.

The difference between paired proportions  is from 64% - 44% = 20%. This mean that media A had 20% more of positive results than media B.

The next step is to comprobe the hypothesis that ther is not difference between paired proportions, taking into account the paired nature of data.