NCJ Number
58643
Date Published
1979
Length
60 pages
Annotation
THIS PAPER CALCULATES THE SAMPLE SIZE REQUIRED TO STUDY RELATIONSHIPS BETWEEN CRIME INCIDENCE AND (1) SURVEY RESPONSES CONCERNING THE VULNERABILITY OF SITES AND (2) FEATURES DESCRIBING THE ENVIRONMENTAL DESIGN OF THE SITE.
Abstract
TO CONTROL FOR THE VARIATION IN CRIME RATE DUE TO SOCIOECONOMIC AND OTHER MACRO CHARACTERISTICS OF THE ENVIRONMENT, CRIME INCIDENCE IS EXAMINED FOR PAIRS OF SITES IN CLOSE PHYSICAL PROXIMITY. THE MATHEMATICAL COMPUTATIONS USED TO DETERMINE THE RELIABILITY OF VARIOUS SAMPLE SIZES ARE GIVEN. GRAPHS ILLUSTRATE THE DIFFERENT PROBABILITY CURVES WHILE TABLES COMPARE BOTH THE EFFECTS OF VARIOUS NUMBERS OF RESPONDENTS ON POWERS OF TEST AND THE POWER OF BINOMIAL AND LOGISTIC TESTS. THESE CALCULATIONS SHOW THAT IF 50 RESPONDENTS RATE EACH SITE PAIR, THE POWER OF THE TEST WILL BE ALMOST AS GREAT AS WOULD BE OBTAINED USING MANY MORE RESPONDENTS. THIS IS TRUE FOR A LARGE RANGE OF ASSUMPTIONS CONCERNING THE NUMBER OF SITES, THE AVERAGE NUMBER OF CRIMES PER SITE, AND THE TRUE RELATIONSHIP BETWEEN SURVEY RESPONSES AND CRIME RATES. FOR INFREQUENT CRIME SUCH AS BURGLARIES IN SINGLE-FAMILY DWELLINGS AND COMMERCIAL ROBBERIES, IT IS DIFFICULT TO OBTAIN SAMPLES WITH AN AVERAGE CRIME COUNT MUCH GREATER THAN TWO CRIMES PER PAIR OF SITES; 40 OR MORE SUCH SITE PAIRS ARE NEEDED TO TEST THE NULL HYPOTHESIS. FOR MORE FREQUENT CRIMES SUCH AS COMMERCIAL BURGLARIES AND BURGLARIES IN MULTIPLE-UNIT DWELLINGS, IF AN AVERAGE OF 10 CRIMES CAN BE OBTAINED IN A SITE PAIR, ONLY 20 PAIRS WOULD TEST THE NULL HYPOTHESIS. FOR FINER DISCRIMINATIONS, MORE SITES ARE NEEDED. (GLR)