Graphics Reference
In-Depth Information
Figure
.
.
Fluctuation diagram of sectors by regions weighted by Total Assets
introuble. hisvariable is bizarrely skewed and sodi
cult tovisualize with standard
plots. In this situation a simple table is better.
ClearlythevariableTL.TAonitsownisveryinformative.hechoiceofthebound-
arylimit(
)isdeterminedbythecontext.Varyingthelimitinteractively usingaslider
confirms that it is a good choice in that the differences in both bankruptcy rates and
the number of bankrupt companies are high.
To investigate the effects of more than one variable, two kinds of parallel plots
can be used with the bankruptcy cases highlighted. Parallel boxplots give univariate
summary comparisons, while parallel coordinate plots potentially offer multivariate
comparisons.
Figure
.
shows the default parallel coordinate plot of the ratios (without Eq.TA
butwith logTA)drawnforalldata exceptforthe
outliers removedinaninitial data
clean. hecompanies that went bankrupt have been selected.heheavy overplotting
may be obscuring information, and the fact that every line is drawn in the same way
whetheritrepresentsonecaseormanymayalsomislead.Nevertheless,somefeatures
can be identified: all bankrupt companies had low values of CL.TA,highvaluesof
Ebit.TA,andmediumvaluesofEbit.Int; there are two unusual bankrupt companies
(one a high outlier on TL.TA, and the other a high outlier on S.TA).
