Introduction

Each point on the chart above is the count of deaths due to hostile fire in Iraq during a seven-day period.  The data are compiled from a database maintained at http://icasualties.org/oif/.  I have included the dates and counts shown on the chart here.

Interpretation

There are three distinct periods with much higher death counts than has been the norm.  I have labeled these with brief descriptions of those periods.  The first such period was during the first phase of active combat, labeled “major combat operations.”  On this chart, that period extends from March 21, 2003 through April 10, 2003.

The second period of markedly higher fatalities is from October 31, 2003 through November 20, 2003.  This period was widely reported at the time as being related to the Muslim holy month of Ramadan, and I have labeled it as such.  The third such period covers from April 2-15, 2004.  During this time, Coalition forces were engaged in battles at Fallujah and against the followers of Moqtada al-Sadr.

Following each high-fatality period is an extended period of time during which the number of fatalities varies randomly about an average value.  The solid red line plotted during each period is the average of the individual weekly counts for that period.  The dashed red line is an upper limit that is calculated from the moving ranges of these same counts.  These charts are control charts or in Dr. Donald J. Wheeler’s terminology, Process Behavior Charts for individual values.  Walter Shewhart developed these charts in the 1930s for industrial quality control purposes.  They are useful tools for analyzing time series.  By applying various “run rules” to such charts, one can evaluate whether the process that produced the data is behaving predictably, or whether it is experiencing unpredictable disruptions.  The three high-fatality periods qualify as disruptions, or events that are likely due to some “special causes.”  The periods of lower fatalities exhibit predictable variation within the range calculated for each period, and that variation is said to be due to “common causes.”  On this chart, points above the upper limit are statistical evidence for a special cause.  I also have examined the charts for other “run rules,” such as 8 consecutive points on the same side of the centerline.

This chart appears to show a steadily increasing rate of fatalities among Coalition forces in Iraq.  The average for the first “quiet” period is 4.8 fatalities per week.  The second average is 7.2 fatalities per week, and the average for the most recent period is 13.8 fatalities per week.  We have reason to believe the periods differ because the data trigger the run rules in several instances when one applies the limits for one period against the data of another.  For example, 11 of the 21 weeks in the most recent period exceed the upper limit (12.4 fatalities in a week) of the April to November 2003 period.  Random fluctuation alone is very unlikely to have caused such a pattern.  We are justified in believing that the environment in Iraq has become more lethal for Coalition forces.

Appendix:  Fatality Count Data