Bureau of Justice Statistics,,,
filename:  cv06myy.csv,,,
Year-to Year Correlation Between Estimates,,,
"data source: National Criminal Victimization Survey, 2006  NCJ  223436",,,
produced by:  Cathy Maston ,,,
authors:  Patsy Klaus and Cathy Maston,,,
refer questions to: askbjs@usdoj.gov  202-307-0765,,,
data of version:  08/29/08,,,

Year-to Year Correlation Between Estimates,,,
"Because of the year-to-year overlap in the sample, the same households and persons contribute to annual estimates for different years.  This year-to-year  correlation between estimates is measured by rho In general:",,,
P equal O when estimates are for the same year,,, 
rho do not equal  O year-to-year comparisons,,,
"When comparing estimates that are 1 year apart, use rho as shown below.",,,
"When comparing estimates that are 2 years apart, multiply rho by 1/2.",,,
"When comparing estimates that are more than 2 years apart, assume rho equal 0.",,,
Following are NCVS year-to-year correlation values for  major crime categories.,,,,
Type of crime,2003-05,2004-05,2004-06,2005-06
,Correlation,Correlation,Correlation,Correlation
Total Crimes,0.2,0.41,0.2,0.41
Total Personal Crimes,0.15,0.3,0.15,0.3
Crime of Violence,0.15,0.31,0.15,0.31
Rape/Sexual Assault,0.02,0.04,0.02,0.04
Robbery,0.02,0.04,0.02,0.04
Assault,0.15,0.3,0.15,0.3
Purse Snatching/Pocket,0.01,0.03,0.01,0.03
Total Property Crimes,0.19,0.38,0.19,0.38
Burglary,0.1,0.21,0.1,0.21
Motor Vehicle Theft,0.04,0.08,0.04,0.08
Theft,0.17,0.34,0.17,0.34
"If estimates are uncorrelated, rho equal 0.  Hence, omitting the term containing rho in the formula will provide an accurate standard error for the difference between uncorrelated estimates.",,,,
"On the other hand, if the two estimates have a strong positive correlation, omitting the last term will cause overestimation of the true standard error. ",,,,
"If the numbers have a strong negative correlation, this will cause underestimation of the actual standard error."