megavideolinks
Joined: 19 Nov 2011
Posts: 273
|
 Posted: Sun Nov 20, 2011 2:27 pm Post subject: Method and Statistical Procedures for the Comparison of Esti |
 |
|
The comparisons in the text have all been tested for statistical significance to ensure that
the differences are larger than those that might be expected due to sampling error. Four types of
comparisons have been made in the text. All statistical comparisons in this report were tested for
significance at the 95 percent confidence level (p<.05), and all reported differences are
statistically significant, unless otherwise noted. One type of comparison tests the significance of
a statistic’s relationship to a specified value, such as 50 percent. When the specified value falls
outside the confidence interval for the statistic, the statistic is said to be statistically different
from the specified value. The confidence interval for an estimate is determined by the following
formula:
CI = x ± t (s.e.)
1
These parameters and their use are explained in U.S. Census Bureau, Source and Accuracy Statement for the
October 2003 CPS Microdata File for Internet and Computer Use in the U.S., Washington, DC, 2004. The general
purpose of the parameters is to adjust the sampling error calculation to account for the complex sampling procedures
used in the Current Population Survey.
48
where x = the estimate for which the confidence interval is desired
t = the critical value, which is 1.96 for a 95 percent interval
s.e. = the standard error of the estimate
For example, the statement on page 27 of this report that “more than half of [singlelocation] Internet users rely solely on home access” was tested as follows:
CI = x ± t (s.e.)
CI = 56 ± 1.96 (1.0)
54 to 58 = 56 ± 1.96
In repeated sampling, the population parameter (i.e., the “true” value) can be expected to
fall within the range of the 95 percent confidence interval 95 percent of the time. In the previous
example, the true value is expected to be between 54 and 58, so the statement in the text that
“more than half of [single-location] Internet users rely solely on home access” is supported. Had
the confidence interval included the range 49 to 53, for example, the statement would not have
been supported because the range includes the value 50.
The second type of comparison tests the hypothesis that there is a linear relationship
between two variables. A bivariate linear regression model is specified, and a statistically
significant regression coefficient supports the hypothesis that there is a linear association
between the dependent and independent variables. To account for the complex sampling
methods of the CPS, a weighted least squares (WLS) regression is used in which the model (e.g.,
the percentage of children who use computers) is weighted by the variance of the dependent
variable (Gujarati 1995).
The third type of comparison tests the statistical significance of the difference of two
statistics. The Student’s t statistic can be used to test the likelihood that the differences between
two percentages are larger than would be expected by sampling error alone.
____________________
real estate in pattaya
dumpster rental cost |
|
