This report includes data from both universe and sample surveys. In the case of universe data, all relevant units are included in the data collection. Thus, there is no sampling error, and observed differences are reported as true. In the case of sample surveys, a nationally representative sample of respondents is selected and asked to participate in the data collection. Since the sample represents just one of many possible samples that could be selected, there is error associated with the sample. To avoid reaching false conclusions about differences between groups or differences over time measured by sample survey data, sampling error is taken into account in statistical tests that are conducted to support statements about differences. Thus, all statements about differences in this report are supported by the data, either directly in the case of universe surveys or with statistical significance testing in the case of sample survey data.
All significance tests of differences in sample survey data are tested at the .05 level of significance. Several test procedures were used, depending on the type of data interpreted and the nature of the statement tested. The most commonly used test procedures were t tests, linear trend tests, and equivalency tests. The t tests were not adjusted to compensate for multiple comparisons being made simultaneously. Trend tests were conducted by evaluating the significance of the slope of a simple regression of the annual data points and a t test comparing the end points. Equivalence tests were used to determine whether two statistics were substantively equivalent by using a hypothesis test to determine whether the confidence interval of the difference between sample estimates was significantly greater or less than a preset substantively important difference. In most cases involving percentages, a difference of 3.0 percentage points was used to determine substantive equivalence or difference. In some tables involving only very small percentages, a lower value was used. The appearance of a “!” symbol (meaning “Interpret data with caution”) in a table or figure indicates a data cell with a high ratio of standard error to estimate (0.30 or greater); therefore, the estimate may be unstable and the reader should use caution when interpreting the data. These unstable estimates are discussed, however, when statistically significant differences are found despite large standard errors.
Although the percentages reported in the tables are generally rounded to one decimal place (e.g., 76.5 percent), percentages reported in the text and figures are rounded from the raw numbers to whole numbers (with any value of 0.50 or above rounded to the next highest whole number). Due to rounding, cumulative percentages may sometimes equal 99 or 101 percent, rather than 100. In addition, sometimes a whole number in the text may seem rounded incorrectly based on its value when rounded to one decimal place. For example, the percentage 14.479 rounds to 14.5 at one decimal place, but rounds to 14 when reported as a whole number. Counts or numbers from universe data are reported unrounded. Estimated counts or numbers from sample survey data are reported rounded to hundreds when they are four- and five-digit numbers and to thousands when they are six-digit numbers.