Most of the statistics in this report are estimates derived from a population. In using a census of an entire population there is not a sampling error, but there is still the possibility of nonsampling error.

Nonsampling errors can be attributed to a number of sources: inability to obtain complete information about all institutions (i.e., some institutions refused to participate, or participated but answered only certain items); ambiguous definitions; differences in interpreting questions; inability or willingness to give correct information; mistakes in recording or coding data; and other errors of collecting, processing, and imputing missing data.

To compensate for nonresponse, adjustments are often made, referred to as imputations. Imputations are usually made separately within various groups of institutions that have similar characteristics. If a particular institution responded in previous and later years, simple interpolations may also be used to substitute for a missing response.

In addition to nonsampling errors, sampling errors may have occurred in the estimates of state per capita personal income, which were used in the simultaneous equation model presented in Chapter IV. Although BEA produced its estimates of state per capita income based on samples, the estimates were aggregated to the state level and the standard errors are likely to be very small. The standard error is a measure of the variability due to sampling when estimating a parameter; very small standard errors signify that a high level of precision would be expected from a particular sample.

Most of the estimates in this report were produced using the SPSS 10.0 software package. SPSS makes it possible for users to specify and generate a variety of univariate and multivariate analyses, including linear regression analysis. In addition, the STATA 6.0 software package was used to produce the estimates for the simultaneous equation models presented in Chapter IV, using the three-stage least squares procedure.