Saturday, February 06, 2010

Liars, Damned Liars and Statisticians

Over the past week, statistics have been raining down on my head. While I do not purport to be an expert in the field of statistics, I do know enough to be dangerous; where and what to question.

Recently a Child Abuse study was released by the US Department of Health And Human Services. Using reports from CPS, LEO (Law Enforcement Organizations), school officials and other 'sentinals', the study estimates there were 743,200 incidences of abuse in 1993 and 553,300 in 2005/06. Further, there were 879,000 estimated incidences of child neglect in 1993, and only 771,700 in 2005/06.

Keep in mind that the source of this data is not strictly from child abuse professionals. This provides a wider and deeper investigation into this social ill. However, including data from individuals provides a bit of variability in the reported data. Some of this variability was filtered through a more or less standardized reporting criteria, but that latitude of variability still exists.

Further, the reporting data was sampled over two, three week periods and then annualized. Two third of the sampled counties were studied from the first week in September through the first week of December. The other one third of the counties were studied from the first week of February through the first week of May. This left, for the most part, these months unreported: December; January; May; June; July; and August.

One of the first things taught in statistics class is to ensure a random, representative sample. Anecdotal evidence as relayed to your's truly, indicate a sharp but brief rise in reports immediately after Christmas and New Years. Neither of these peaks are included in the sample periods.

Further, using the 2008 Quarterly Domestic Violence Report from the Nevada Department of Public Safety, the peak months of reported DV is April, May, June, July, August and September. National monthly or quarterly statistics were not readily available, so, this will need to suffice.

See any issue? The federal report missed several months of high reported abuse.

Further, other government studies show different numbers. According to the CDC, there were 905,000 childhood victims of maltreatment from October, 2005 through September, 2006 as reported by state and local child protective services. According to the Health and Human Services report, a total estimated of 1,256,600 occurred. This latter number included an estimation of official and unofficial reports. So, if there were 905,000 official reports, unofficial reports would constitute 351,600. Sort of low, in my opinion.

To make things just a little stranger, the US Census Bureau reported there was a total of 1,860,259 reports of child abuse and neglect that were investigated in 2007. This number blows the Health and Human Services report out of the water.

Either way, if that number is greater than one, it is too many.

1 comment:

Ann said...

The issue is, the report was released by the DOH. That's the Doh! department of our government. I bet if you called them on their shoddy reporting style, they would say, "Doh!" and move on!

Thanks for the post, though. If these results are skewed, just imagine what the real numbers are. Disgusting, that's what.