GLOSSARY AND TECHNICAL NOTES
Note: In the reports delivered to clients certain words are italicized and underlined.
These words are discussed in a Glossary provided with the report. The Glossary is also replicated below.
Associated weights. Also see stepwise multiple regression and prediction formula. In the final prediction formula each significant demographic variable is weighted in proportion to the contribution it makes to the predicted (i.e., expected) value of volume-of-business.
Business type. By business type we mean the 48 businesses listed on our web site and order form. These are extrapolations from the Consumer Expenditure Survey, and in some cases, are combinations of two or more categories from those surveys.
Demography. In this report, demography refers to social statistics as collected from several primary sources: a) the most recent ten-year census, b) estimates and projections for the years following that census, c) the US Department of Commerce studies of industries of 2004, and d) household spending patterns derived from annual Consumer Expenditure Surveys
Demographic variables. By demographic variables we mean those social variables that are statistically correlated with consumer expenditures. Several hundred variables were considered initially. After repeated multiple regression trials, we condensed these variables to ten measures. They included one or more indices of income, household size, number of household units, age, population density, growth in population, etc. We also included one geographic variable which, in a few cases, correlated with consumer expenditure: number of square miles in the geographic area of interest (county or ZIP).
Order of importance. Also see stepwise multiple regression and significant variables. Demographic variables are brought into the final prediction formula in a particular order. That order is determined by the contribution each makes to explaining the variance of volume-of-business. The order is from high to low and each has a calculated measure of its importance (T value). This report sorts the variables according to their T value.
Predict. See stepwise multiple regression below.
Prediction formula. The stepwise multiple regression yields a prediction formula. That formula consists of a constant plus a series of terms. Each term represents one significant demographic variable multiplied by its associated weight. When the entire formula is evaluated numerically, it yields the "expected value" as given in this report.
Significant variables. In the context of this report, a demographic variable is said to be significant if it makes at least a specified level of contribution to explaining the variance in volume-of-business for the business type selected. See variance below.
Stepwise multiple regression. The multiple regression procedure is designed to construct a statistical model describing the impact of two or more independent variables, such as demographic variables, on one dependent variable, such as volume-of-business for a particular business type. A stepwise regression finds the subset of the demographic variables that best explains volume-of-business. Thus, the final model will be different for every business and every type of geography. Further, the final model may be used to make a prediction, which in our report is referred to as the “expected value."
Variance. The volume-of-business for any one business type will, of course, vary from one ZIP or County to another. Although variance has a mathematical definition it is, at root, a measure of the degree to which any given variable "spreads out," or varies, from its mean. The goal of regression analysis, as used for this report, is to explain, to the greatest extent possible, this tendency to vary.
Volume-of-business measures. We used volume-of-business as a synonym for consumer expenditure because the focus of our reports is on selling as opposed to buying.
ZIP-code level. Demographic information from census reports is only available in units considerably smaller than ZIP codes. In order to obtain demography for a ZIP-code area, several or many such "census blocks" are combined. Unfortunately the boundaries of a group of census blocks do not exactly correspond to the boundaries of a ZIP-code area. Thus, the census blocks that lie along the boundary of the ZIP are apportioned between the ZIP of interest and the adjacent ZIP. So, when we refer to ZIP-code level, we mean a close approximation based on the combining and splitting of census block data.