**Controls Over Accountable Property at the Baltimore Field Division of the Federal Bureau of Investigation**

**Report Number 04-37
September 2004
Office of the Inspector General**

Appendix 8

Statistical Sampling Design and Computation of Estimates

The statistical sampling universe consisted of 11,691 elements (Total number of Property Items.) This universe was composed of two sub universes: (a) "Under-$1,000 Property" of 6,071 elements, and (b) "Over-$1000 Property," or accountable property, containing 5,620 elements. Two different sampling designs were employed for these two sub-universes. When testing, each element was tested for a characteristic, which we term as a discrete variable involving nominal measure (refer to the body of the Audit Report for a description of the audit tests conducted).
We employed a stratified random sampling design to provide effective coverage and to obtain precise estimates of the test statistics for the under $1,000 inventory. We estimated totals for two types of exceptions in this sub-universe. Our conservative estimates of lower limits, on 95 percent confidence intervals for the test statistic of total number of type 1 exceptions ("could not locate property") is 570 and that of type 2 exceptions ("could not determine") is 133. The estimates were computed using the formula (A) given below. The notation and the formulae are as follows: [
We employed a random sampling design for testing and to obtain estimates of the test statistic of the accountable property sub-universe. We estimated totals for one type of exceptions in this sub-universe ("could not determine"). Estimate of lower bound on 95 percent confidence interval for the test statistic of total number of exceptions is 17. The estimate was computed using the formula (B) given below. The notation and the formulae are as follows: [ |