Field Test Evaluation of Conservation Retrofits of Low-Income, Single-Family Buildings in Wisconsin: Audit Field Test Implementation and Results Page: 47 of 84
This report is part of the collection entitled: Office of Scientific & Technical Information Technical Reports and was provided to Digital Library by the UNT Libraries Government Documents Department.
The following text was automatically extracted from the image on this page using optical character recognition software:
The large confidence intervals associated with the savings estimates may
make it difficult to act decisively on the basis of the results of this study.
The simplest way to produce more precise estimates is to use a larger sample,
however, there are usually practical constraints on how large a sample can be
used. The sample size must be quadrupled in order to halve the size of the
An examination of the variances displayed on Table 4.3 suggests two other
approaches. One approach is to reduce the variances associated with individual
measurements. These variances are the sum of the individual regression vari-
ances. A more stringent data quality criterion, such as a minimum regression R2
of 0.9, would be expected to reduce the variance from this source by eliminating
from the sample some of the houses with larger standard errors. This process,
with a minimum R2 of 0.85, was tested on the control houses. The criterion
eliminated four houses and reduced the variance contributions of the individual
measurements by more than one-half.* The criterion was not adopted because of
its after-the-fact nature and because eliminating this source of variance would
not greatly reduce the standard error of the mean for the reportable savings
The largest source of variance is the contribution due to dispersion of the
audit group NAHSs. As discussed previously, this dispersion of NAHSs was the
expected result of the ADRP. It should also be noted that the variance calcu-
lated for the audit predictions (last column of Table 4.3) and the variance from
the dispersion of measured audit group NAHS (first column) are nearly the same.
The only way to substantially reduce this component of variance is to study a
more homogeneous group. For instance, this source of variance might be nearly
*The variance caused by the dispersion of control group NAHSs was reduced
almost as much as the variance caused by individual measurements, suggesting
that much of the variance of the control group savings is due to measurement
errors, not actual changes in energy consumption patterns between the pre- and
post-retrofit period. It also suggests we may be double counting contributions
to the variance of the control group. If so, the error is small. Eliminating
either source of variance would reduce the standard error of the mean savings by
less than 10%.
Here’s what’s next.
This report can be searched. Note: Results may vary based on the legibility of text within the document.
Tools / Downloads
Get a copy of this page or view the extracted text.
Citing and Sharing
Basic information for referencing this web page. We also provide extended guidance on usage rights, references, copying or embedding.
Reference the current page of this Report.
McCold, L.N. Field Test Evaluation of Conservation Retrofits of Low-Income, Single-Family Buildings in Wisconsin: Audit Field Test Implementation and Results, report, January 1, 1988; United States. (digital.library.unt.edu/ark:/67531/metadc740901/m1/47/: accessed November 21, 2018), University of North Texas Libraries, Digital Library, digital.library.unt.edu; crediting UNT Libraries Government Documents Department.