Table III. Data, analysis and results of... | Emerald Publishing

Table III.

Data, analysis and results of an innovative CRA application by Williamson et al. (2004)

Data	Analysis	Group	Result
521 incident reports (c.f. system failures) in IT services managed by a global corporation, with up to 12 pages each, categorized by three former locations and written by experts	Calculation of the average influence of each word across the reports	Influence	Most influential words in the reports
	Categorization of each word based on an Ontological Dictionary developed by IT experts	Text	Number of influential words (out of the 500 most influential, on average) by category
	Comparison and contrast of words’ average influence, by location and seller	Influence	Locations and sellers more/less influential in reports
	Exploratory Factor Analysis by Principal Components with Varimax rotation for the set of reports, based on the influence values of the words that occurred in at least 10% of the reports	CRA + other methods	Main underlying factors for words, interpreted as key themes in the set of reports
	Calculation of the influence of a theme in a report as the average of the influences of the words associated with it (the theme) in that report	Influence	Influence of each theme for each report
	Analysis of the stability of the influence of themes over time by Statistical Process Control Charts, using the influences of a theme over a period of 1 month (i.e. in 3 to 10 reports) as a subgroup	CRA + other methods	Stability/Significant variations of themes’ influence over time
	Time Series Analysis, by analyzing significant correlations between the influence of themes whose occurrences were not too much dissociated over time (i.e. from 0 to 3 months away)	CRA + other methods	Inferences of possible causal relationships between themes, consolidated in a causal map, in which: nodes represent significantly correlated themes; arrows indicate the inferred causal direction; and their values means the elapsed time between themes` occurrences (a format that enables the differentiation between root causes and symptomatic effects) [Knowing that correlation does not imply causation]

Data	Analysis	Group	Result
521 incident reports (c.f. system failures) in IT services managed by a global corporation, with up to 12 pages each, categorized by three former locations and written by experts	Calculation of the average influence of each word across the reports	Influence	Most influential words in the reports
	Categorization of each word based on an Ontological Dictionary developed by IT experts	Text	Number of influential words (out of the 500 most influential, on average) by category
	Comparison and contrast of words’ average influence, by location and seller	Influence	Locations and sellers more/less influential in reports
	Exploratory Factor Analysis by Principal Components with Varimax rotation for the set of reports, based on the influence values of the words that occurred in at least 10% of the reports	CRA + other methods	Main underlying factors for words, interpreted as key themes in the set of reports
	Calculation of the influence of a theme in a report as the average of the influences of the words associated with it (the theme) in that report	Influence	Influence of each theme for each report
	Analysis of the stability of the influence of themes over time by Statistical Process Control Charts, using the influences of a theme over a period of 1 month (i.e. in 3 to 10 reports) as a subgroup	CRA + other methods	Stability/Significant variations of themes’ influence over time
	Time Series Analysis, by analyzing significant correlations between the influence of themes whose occurrences were not too much dissociated over time (i.e. from 0 to 3 months away)	CRA + other methods	Inferences of possible causal relationships between themes, consolidated in a causal map, in which: nodes represent significantly correlated themes; arrows indicate the inferred causal direction; and their values means the elapsed time between themes` occurrences (a format that enables the differentiation between root causes and symptomatic effects) [Knowing that correlation does not imply causation]

Source: The authors