Often thought of as qualitative data, anything produced by the interpretation of laboratory scientists (as opposed to a measured value) is still a form of quantitative data, albeit in a slightly different form. As such different statistical methods from those used for data routinely assessed in the laboratory are required.
The approach taken will vary depending on the nature of the assessment being made, and the result being produced/reported. Analysis falling into each of these categories, and their application to uncertainty quantification/reliability of results is summarised below. The report provided will clearly outline which methods have been used along with an interpretation of results provided. In depth discussion of each of the methods outlined below is out with the scope of this protocol. The summary table below indicates in what situation each may be used.
|Cohen’s Kappa Statistic||Categorical data – two observers|
|Fleiss’s Kappa and Conger’s Kappa||Categorical data >2 observers|
|Weighted Kappa||Ordinal data|
|Weighted Kappa with factors||Ordinal data using Factors|
|Intraclass correlation coefficient||Continuous data|
Statistical methods employed for the assessment of inter observer reliability