Center (NPC) in Jeffersonville, IN. Telephone 
follow-up from a NASS Data Colleetion Center 
began April 2014 to nonrespondents who were 
mailed a report form from NPC. 
Data were collected for a select group of operations 
by the NASS field offices. To minimize the number 
of agency contacts, operations were included in this 
group if they were scheduled for contact by NASS 
for other agricultural surveys. Report forms were 
labeled at NPC and sent to the field offices in 
December 2013. Field office staff collected data by 
personal enumeration or by phone from January 
2014 through May 2014. For a description of the 
adjustment for nonresponse, see Estimation. 
REPORT FORM PROCESSING 
Data Capture 
All report forms returned to NPC were immediately 
checked in, using bar codes printed on the mailing 
label, and this check-in process removed them from 
follow-up mailings. All forms were reviewed prior to 
data keying to identify inconsistencies and ensure 
that the data could be keyed. Major inconsistencies, 
respondent remarks, blank report forms, and large 
irrigation cases were reviewed by analysts and 
adjusted prior to data keying as needed. All forms 
with any data were scanned and an image was 
created for each page of a report form. 
Data Editing and Anaiysis 
Data from each report form were processed through 
a computer edit which flagged inconsistent entries. 
Each flagged entry was reviewed by staff. In some 
cases, respondents may have failed to provide all of 
the information requested, only indicating the 
presence of an item but not the amount. Eor those 
data that would not be machine imputed they were 
estimated by the analyst based on other responses in 
the geographic area and by similarly sized farms. 
After the initial edit, an imputation program supplied 
missing data and made adjustments based on 
responses of similarly sized farms within the same 
geographic area. Data entries of large magnitude and 
data items that were changed significantly in the 
computer edit process were reviewed and verified by 
analysts. 
A - 2 Appendix A 
Prior to publication, tabulated totals were reviewed 
to identify and resolve remaining inconsistencies and 
potential coverage problems. Comparisons were 
made with 2012 census data, 2008 Farm and Ranch 
Irrigation Survey data, and other available check 
data. The data were processed through a disclosure 
program to prevent data from being published that 
could be sourced back to an individual operation. 
Imputation 
After the initial edit, imputations were made for 
missing data on quantity of water applied, well and 
pump characteristics, energy cost of well pumps, 
individual crop yields and quantity of water used, 
horticulture water sources, maintenance and repair 
costs, and expenditures listed in report form section 
15. 
ESTIMATION 
Data were summarized for the Nation as a whole, for 
each of the 50 States, and for the geographic 
domains known as Water Resources Regions (WRR) 
(see Appendix B for detailed description). The 
estimation methodology consisted of two weighting 
components that made up the total FRIS weight. The 
first component was the fully adjusted weight pulled 
in from the 2012 Census of Agriculture. This weight 
accounted for any list incompleteness and 
undercoverage from the 2012 census. The second 
component was the sampling rate used for the FRIS. 
This expansion factor was the inverse of the 
selection probability for the sample farms in a 
stratum. This expansion factor was reweighted at the 
stratum level to account for whole-farm 
nonresponse. The nonresponse adjustment factor 
used to reweight the expansion factor was the ratio 
of the number of sample farms in a stratum to the 
number of sample farms that responded to the survey 
in that stratum. The assumption underlying this 
weighting approach to survey nonresponse was that 
survey respondents and nonrespondents within a 
stratum constitute a homogeneous population, thus 
allowing respondents to represent nonrespondents. 
An expanded data value for a sample record was 
obtained by multiplying the data value by the total 
FRIS weight. State totals for a characteristic were 
estimated by summing the expanded data values 
from all responding sample records across all strata 
within the State. National estimates were obtained by 
2012 Census of Agriculture 
USDA, National Agricultural Statistics Service 
