NASS Review of Survey

NASS Review of OMB No. 0581–0310 (previously 0580-0018)

AMS - Survey of Customers of the Official Grain Inspection and Weighing System

February 19, 2019

Review conducted by Jared Pratt (Mathematical Statistician) at USDA-NASS Summary Estimation Disclosure Methodology Branch (SEDMB)

General Comments:

Good three year plan on grain inspection survey. Defines population, expected response rates, and plan. The goal is to gather their customers’ opinions of their system changes.

Suggestions

It seems like there are two sub-populations, a defined number of known customers and an undefined number of non-customers. The sample is for the known customers and the data collection is for both populations. The overall population is all visitors to the FGIS Primary/Secondary websites.

Some stratum have a low sample size and therefore estimates based on that stratum should be avoided. Beware of disclosure of small sample sizes within each stratum. It is unclear if your final published estimates are at the population level or at the stratum level.

Consider using the “Finite Population Correction Factor” since your sample of 1,100 is more than 5% of the population of 9,000.

The sampling frame is created by FGIS from email, trade publications, agency newsletters, and trade/stake holder meetings. Lists are inherently incomplete, so consider making adjustments to the final estimates due to the incompleteness of your list.

The link on the website should help the response rate. It seems there should be an identifier to connect the sample to the respondent. Also, are responses allowed outside of the designated sample? If so, that estimate should be a classified as a non-probability estimate.

If sampling weights are assigned to each of your 1,100 samples, then consider using a non-response adjustment, such as reweighting, for the survey indications and standard errors.