Appendix E.3 Comments from the National Agricultural Statistics Service

APPENDIX E.3

Comments from the National Agricultural Statistics Service

REVIEW OF FNS CHILD NUTRITION PROGRAM OPERATIONS STUDY II (CN-OPS-II)

Review by Evan Schulz (Mathematical Statistician), USDA NASS

Supporting Statement A

• Item A.2, first paragraph: “The statistics will be calculated using weighted data, such that all results will provide national estimates.” Specifics on weighting [maybe in SSB]?

• Item A.2, directly after “Frequency of information collected” heading: Consider changing “All CN directors and sampled SFAs will complete…” to “All CN directors and sampled SFAs will be asked to complete…” There will inevitably be some nonresponse, at least in the survey of SFA directors (which you indicate in the burden table in Item A.12). Similarly, consider changing the third sentence to say, “Because the CN Director Survey is a census of State directors, they will be asked to participate every year.”

• Item A.5: Consider including how you define a small entity, and possibly how that definition was chosen.

• Item A.8, paragraph 3: Should “… focusing questions on current CN programs polices…” read “… focusing questions on current CN program policies…”?

• Item A.12, paragraph 1: Fix the burden hours in the statement, “The total annualized burden to the public is 4,098.9.8…” The figure should be 4,098.9.

• Item A.12, tables: Each number in the “Total Annual hour burden” column in Table A.2 (Annualized Cost to Respondents) should agree with the ‘All: Total Annual hour burden’ column in Table A.1 (Estimates of Respondent Burden). They do not agree. In particular, the figures in Table A.2 should include burden to non-respondents.

Supporting Statement B

• Item B.1, page 4: There is a paragraph that begins with “ar”. Remove this.

• Item B.2, under the “Estimation and Calculation of Sampling Errors” heading: This section discusses constructing weights to adjust for nonresponse – specifically, using predictors to model response propensity. Consider mentioning briefly what kind of model will be used (e.g. logistic regression, classification trees, etc.)