1. It is not clear to me if the estimates are intended to produce State level and regional estimates, as well as national estimates. In SSB, it seems the sampling stratification would sort by state and then size of firm (suggesting (maybe) that State level estimates are desired). Please clarify.
Response: State level estimates are not desired. We will produce estimates only at regional and national levels. We had planned to sample 3000 bison producers, but because results from the 2012 Census of Agriculture indicate that there now are fewer than 3000 bison producers in the United States, we will send the questionnaire to all producers on the NASS listframe in July 2014 (expected to be about 2800) and the study will be a census. Consequently, there will be no sampling stratification.
2. We see that the projected response rate is 20%. We also see that no telephone follow up is planned of initial nonrespondents, or subsets of nonrespondents. Was telephone follow up to improve response rates considered? Was selective telephone follow-up of nonresponders considered as a way to better model (and adjust for) potential nonresponse bias?
Response: Telephone followup of nonrespondents was discussed but was considered not to be cost-effective because of the nature of this study. This is the first time U.S. bison producers are being surveyed by NAHMS, and we are working with Bison industry associations and other entities to promote the study and encourage participation. The projected response rate of 20% is based on NASS’ experience with industries that are routinely surveyed; the actual response rate might be higher for an industry that has not been surveyed before. Control data from the NASS listframe will be used to evaluate any nonresponse bias.
3. Thank you for presenting the power estimation method and results. It seems that the sample size is powered to produce point estimates with precision, but it is not clear to me if the sample size is powered for comparison of estimates (for example, differences in bison farming methods by size of firm). Please confirm.
Response: The sample size is not powered for comparison of estimates. One of our goals is to develop descriptive statistics with a coefficient of variation (CV) less than 20 percent. This applies to any sub-population estimates (regional, operation-type, size category) as well as national estimates. However, the primary goal of NAHMS studies is to develop accurate national estimates. In setting the sample size, we try to achieve a balance of precision, cost, and producer burden.
| File Type | application/vnd.openxmlformats-officedocument.wordprocessingml.document |
| Author | Parker, Margaret A - APHIS |
| File Modified | 0000-00-00 |
| File Created | 2021-01-28 |