Thereafter an overall check will be done using frequency tables to detect unusual values, inconsistencies and missing values.
The second stage of quality control was performed centrally at the Estonian site, by double-checking each local dataset, attempting to detect other errors, such as incompleteness (missing values), inconsistencies (incorrectly followed skip-outs), irregularities (numbers inserted in text variables), and out of range data (i.e. very large number of siblings or sexual partners).
The principal drawbacks of stepwise selection include biases in parameter estimation and reported p-value, inconsistencies among model selection algorithms, an issue of multiple hypotheses testing, and the possibility of missing the optimal model.
When the C---/X↔Y method was used, the values of inconsistency varied between 0.59% and 3.25%.
When the ChsI/X↔Y method was used, the values of inconsistency are much better than in cases in which the identity of the InChI component is required.
If the only requirement was that the linked entities have compatible InChI identifiers (i.e., the CHsI/X⇔Y method was used), the values of inconsistency dropped significantly in many cases.
Data are routinely checked for completeness, out of range values and inconsistency [ 18].
(Also see table 5 in web extra appendix 3 for 95% credibility intervals and P values of inconsistency estimates).
The values of inconsistency coefficient (I) of both sensitivity and specificity of the halo sign, showed significant heterogeneity concerning the results between studies.
The AC1 statistic can take any value from -1 to 1, with 1 indicating full consistency, 0 absence of consistency (the degree of agreement expected by chance), and negative values indicating inconsistency or disagreement.
