THE TIME-VARYING NATURE OF CRIME DATA IS OFTEN OVERLOOKED RESULTING IN MISLEADING OBSERVATIONS AND FORECASTS. REPORTS AND OTHER PRIMARY SOURCES ARE GIVEN. THE LIMITATIONS OF FOUR LEVELS OF ANALYSIS ARE DETAILED: ELEMENTARY STATISTICAL METHODS, PRELIMINARY ANALYSIS, DETAILED ANALYSIS, AND STOCHASTIC MODEL ANALYSIS. OF THESE, ONLY THE STOCHASTIC MODEL HAS GOOD PREDICTIVE CAPABILITY, ALTHOUGH THIS CAPABILITY IS LIMITED BY THE TYPE OF DATA HANDLED. ONE OF THE MOST RELIABLE, ALBEIT ONE OF THE MOST COMPLEX, APPROACHES TO DATA IS THE AUTOREGRESSIVE INTEGRATED MOVING AVERAGE (ARIMA) METHOD. MATHEMATICAL TECHNIQUES USE TO BUILD THIS MODEL ARE ILLUSTRATED. VARIOUS ARIMA SOFTWARE PROGRAMS AVAILABLE FROM UNIVERSITIES AND COMMERCIAL SOURCES ARE DESCRIBED WITH STRENGTHS AND WEAKNESSES DISCUSSED. ARIMA PROGRAMS HAVE HIGH PREDICTIVE RELIABILITY FOR ENGINEERING DATA. THEIR USE FOR CRIME STATISTICS IS MORE LIMITED BECAUSE OF THE RANDOM NATURE OF MANY CRIMES; HOWEVER, THEY PROVIDE GOOD ANALYSIS OF TRENDS. AN ALTERNATIVE PROGRAMS WHICH DEALS WITH CYCLIC TRENDS IS THE CENSUS BUREAU X-11, WHICH IS ALSO DESCRIBED. PRACTICAL SUGGESTIONS ARE GIVEN FOR ITS USE. TABLES ILLUSTRATE VARIOUS STATISTICAL MANIPULATIONS POSSIBLE WITH BOTH METHODS, PREPARATION OF DATA FOR ARIMA ENTRY, ANALYSIS OF OUTPUT, AND SAMPLE MODELS GENERATED FROM BOTH ACTUAL AND SIMULATED DATA. APPENDIXES GIVE A SUMMARY OF ARIMA SOFTWARE NEEDS, SAMPLE PROGRAM STEPS, AND A BIBLIOGRAPHY IS INCLUDED.