Pie charts or bar charts or cross-tables give some useful results for
safety measures to be developed, but these are not sufficient under
situations when observations do not pertain to the root underlying
causes and the symptomatic or manifest variables observed are
categorical in nature. Under such circumstances, cross-table data
needs more in-depth study and correspondence analysis (CA) plays
a significant role. CA helps from several counts: (i) proving row
variable’s contribution, (ii) providing column variable’s contribution, (iii) providing association between row and column variables,
and (iv) extracting hidden or underlying causes or dimensions.
While the first two issues can also be addressed through charts
or plots (e.g., bar, pie) and the third one through v2
statistic,
respectively, but the fourth one be achieved through CA only.
The CA based analysis of derailment data extracts four impor-tant rules for the steel plant studied. They are reproduced below
for discussion.