1 / 33

For-Hire Survey

For-Hire Survey Survey Design Recommendations Presented by Jim Chromy jrc@rti.org NRC 2006 “For-Hire” Concerns More like commercial sector Estimation does not recognize design Physical, financial, and operational constraint biases Fish caught and not brought to dock

Download Presentation

For-Hire Survey

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. For-Hire Survey Survey Design Recommendations Presented by Jim Chromy jrc@rti.org

  2. NRC 2006 “For-Hire” Concerns • More like commercial sector • Estimation does not recognize design • Physical, financial, and operational constraint biases • Fish caught and not brought to dock • Cover small and private landing points • Dual frame to reduce bias: logbooks

  3. Themes • Survey design is often intuitive. • Theoretically sound design depends on specific procedures for sampling and estimation • Many acceptable solutions • None will be perfect

  4. Topics • General survey vs. fisheries survey terms • Probability sampling procedures at all stages • Sample size to meet analytic needs • Sample allocation to control sampling error • Estimation based on sample design, including appropriate weighting. • Coverage and response issues

  5. Before Sampling • Conceptual population • Points of departure or area fished • Vessels • Anglers • Catch • Conceptual domains • Region • Catch species • Time periods

  6. Sampling Frames • Try to cover conceptual population • List of labels and rules • Labels are unique and of finite number • Rules are links to actual population elements—e.g., names and contact information for vessels • Labels can be selected using probability sampling. • Rules permit identification of the sample.

  7. Frame Examples • Directory of for-hire vessels operating from NC coast during a specified period • List of for-hire fishing trips returning to a single landing during a specified time period • List of anglers participating in a vessel trip; stringer tags plus list of unsuccessful anglers • Order number for fish landed by an angler: could be ordered by size

  8. Frame Structure • Simplest: list • Example: for-hire vessel directory for NC • Used for telephone survey component • Multi-stage or nested lists • Landing area by time period • Vessel trips ending in above • Anglers aboard a vessel trip • Fish landed by an angler • Crossed frames: spatial vs. temporal

  9. Temporal Frame Structure • Year • Month: 1, 2,…,12 • Week ending on Sunday: 1, 2,…,52 • Kind of day:1=weekend, 2=weekday • Day: (Sat, Sun) (M, T, W, Th, F) • Hourly periods including night time:(

  10. Sample Size • Must be adequate to meet analytic needs • No 10 or 20 percent sampling unless those rates are justified by need and adequacy to meet that need. • May be limited by budget

  11. Stratification of Frames (1) • For administrative control • For workload distribution • For analytic purposes—match domains • To allow different sampling rates • To identify certain exclusions—reduce coverage in a controlled manner

  12. Stratification of Frames (2) • To increase efficiency, reduce sampling error and control costs! • To permit different sampling and data collection methods by strata: e.g., dockside vs. at-sea.

  13. Example: For-Hire Directory • State • Region within state • Headboat vs. charter • Capacity in anglers • Active status for survey period: e.g., active, verified as inactive, not sure. • Need to know number of vessels in each stratum and their labels.

  14. Example: Vessel Trips • Landing area • Time period of landing • Order of landing • Vessel capacity • Need to know number of vessel trips in each stratum and their labels • Label could be order of landing during specified time period

  15. Example: Anglers • May take all on small vessels: each angler selected with probability 1.0 • Large vessels intercepted at dock (sample size may be determined by time available) • At-sea observation on large vessels

  16. Intercepted At Dock • Frame problem—list or order of departure • If time permits, pre-identify some anglers for sampling with probability 1.00 (based on species caught, size, or other factors) • Sample remainder at lower rate or rates • Include all anglers in an assigned stratum • For each stratum, know N, n, and probability of selection (n/N).

  17. At-Sea • Sampling for discard observation • Frame in time and location on vessel • Mark locations (areas along rail) and sample by time period once fishing begins. Observe and record all discards. • Sampling for retained catch at completion of fishing • Similar to intercept problem • More time to obtain data

  18. Stratified Sample of Angler’s Landed Catch • Classify fish into groups/strata • Rare species • Size • Record number of fish in each stratum • Select probability sample by stratum • For each stratum, record N, n, and sampling rate (n/N) • Simplest case: “take all”

  19. Probability Sampling Methods • Simple Random Sampling • Systematic Random Sampling • PPS Sampling • All can be applied within strata • All can be applied at various stages of sampling

  20. Estimation • General topic for another team • Must be based on design • General form: weight inversely to selection probability • Weights may be adjusted for nonresponse or undercoverage

  21. Selecting a Simple Random Sample • All samples of size n have equal probability • Each unit is selected with probability n/N • Estimation weight: W=N/n. • Random permutation is easy to apply: currently used for telephone survey samples

  22. Random Permutation Example

  23. Simple Random Sample of 7 of 25

  24. Random Sample of 7 out of 12

  25. Systematic Random Sampling • Select a random number between 1 and k for first sample label • Then select every k-th label to end of list • Probability of selection is 1/k. W=k. • Nice if N=nk. • Alternatives for non-integer k, i.e. k=N/n.

  26. PPS Sampling • Many acceptable methods • SAS/STAT Proc Surveyselect • Several methods available • My favorites: Method=Chromy • Output provides probability of selection, P • Weight = 1/P.

  27. Sample Allocation • Achieved through stratification and sample allocation • Can also be achieved through PPS sampling. • Improve precision • Control costs • Fishing pressure is a natural size measure or basis for sample allocation.

  28. Form of Estimates Total effort Average CPUE

  29. Nonresponse Adjustments • Weight adjustment for unit level nonresponse • Imputation for partial nonresponse

  30. Poststratification • Ratio-type adjustments to incorporate known (or better) data for related statistics • Can help adjust for undercoverage • Basis for adjustment should be justified and re-evaluated on a regular basis. • Can also adjust for unusual sample outcome. • After sampling stratification and adjusted estimation

  31. Double Sampling • Technique for adjusting biased estimates perhaps based on low cost approach • Uses smaller (high cost) sample to fine-tune. • Example: 100 percent logbook data could be adjusted based on dockside or at-sea samples for a sample of vessel-trips.

  32. Many Techniques Available • Ultimate approach will be a mix of methods • Tough problems remain. • Continuous improvement plant should begin.

  33. Thank You

More Related