| Data | To correct for these multiple comparisons we apply the Benjamini-Hoch-berg procedure [17] to control for the false discovery rate a. |
| Data | For example, if 100 comorbidities are identified with a false discovery rate a of a = 0.01, the eXpected number of false positives among these comorbidities is one. |
| Data | We Will therefore be interested in the recall R(a) as a function of the false discovery rate a. R(a) is the probability that a diabetic comorbidity listed in Table 1 is also identified by our co-occurrence analysis at a given level of a. |
| Results/Discussion | Each diagnosis where the null hypothesis of statistical independence with either DM1 or DMZ can be rejected with a given value of the false discovery rate in at least one of the age groups is identified as a comorbidity. |
| Results/Discussion | A false discovery rate of a = 0.001 gives a list of 75 significant comorbidities and a recall of R(a = 0.001) = 0.59. |