Some Insights into Data Weighting in Integrated Stock Assessments André E. Punt 21 October 2015 Index-1 length-4.

Slides:



Advertisements
Similar presentations
Modeling Recruitment in Stock Synthesis
Advertisements

An exploration of alternative methods to deal with time-varying selectivity in the stock assessment of YFT in the eastern Pacific Ocean CAPAM – Selectivity.
Multiple Regression Analysis
FTP Biostatistics II Model parameter estimations: Confronting models with measurements.
Gavin Using size increment data in age-structured stock assessment models CAPAM growth workshop: Nov 2014.
Modeling fisheries and stocks spatially for Pacific Northwest Chinook salmon Rishi Sharma, CRITFC Henry Yuen, USFWS Mark Maunder, IATTC.
Estimating Growth Within Size-Structured Fishery Stock Assessments ( What is the State of the Art and What does the Future Look Like? ) ANDRÉ E PUNT, MALCOLM.
An Overview of the Key Issues to be Discussed Relating to South African Sardine MARAM International Stock Assessment Workshop 1 st December 2014 Carryn.
Growth in Age-Structured Stock Assessment Models R.I.C. Chris Francis CAPAM Growth Workshop, La Jolla, November 3-7, 2014.
Model Selection for Selectivity in Fisheries Stock Assessments André Punt, Felipe Hurtado-Ferro, Athol Whitten 13 March 2013; CAPAM Selectivity workshop.
C3: Estimation of size-transition matrices with and without molt probability for Alaska golden king crab using tag–recapture data M.S.M. Siddeek, J. Zheng,
The current status of fisheries stock assessment Mark Maunder Inter-American Tropical Tuna Commission (IATTC) Center for the Advancement of Population.
Econ Prof. Buckles1 Multiple Regression Analysis y =  0 +  1 x 1 +  2 x  k x k + u 1. Estimation.
458 Fitting models to data – II (The Basics of Maximum Likelihood Estimation) Fish 458, Lecture 9.
Statistics for Managers Using Microsoft Excel, 4e © 2004 Prentice-Hall, Inc. Chap 7-1 Chapter 7 Confidence Interval Estimation Statistics for Managers.
Basic Business Statistics, 10e © 2006 Prentice-Hall, Inc. Chap 8-1 Chapter 8 Confidence Interval Estimation Basic Business Statistics 10 th Edition.
Evaluating Hypotheses
Efficient Estimation of Emission Probabilities in profile HMM By Virpi Ahola et al Reviewed By Alok Datar.
Hui-Hua Lee 1, Kevin R. Piner 1, Mark N. Maunder 2 Evaluation of traditional versus conditional fitting of von Bertalanffy growth functions 1 NOAA Fisheries,
458 Fitting models to data – III (More on Maximum Likelihood Estimation) Fish 458, Lecture 10.
Lehrstuhl für Informatik 2 Gabriella Kókai: Maschine Learning 1 Evaluating Hypotheses.
458 Fitting models to data – I (Sum of Squares) Fish 458, Lecture 7.
Statistics for Managers Using Microsoft Excel, 5e © 2008 Pearson Prentice-Hall, Inc.Chap 8-1 Statistics for Managers Using Microsoft® Excel 5th Edition.
Statistics for Managers Using Microsoft Excel, 4e © 2004 Prentice-Hall, Inc. Chap 7-1 Chapter 7 Confidence Interval Estimation Statistics for Managers.
The (potential) value and use of empirical estimates of selectivity in integrated assessments John Walter, Brian Linton, Will Patterson and Clay Porch.
by B. Zadrozny and C. Elkan
Basic Business Statistics, 11e © 2009 Prentice-Hall, Inc. Chap 8-1 Chapter 8 Confidence Interval Estimation Basic Business Statistics 11 th Edition.
Confidence Interval Estimation
Population Dynamics Mortality, Growth, and More. Fish Growth Growth of fish is indeterminate Affected by: –Food abundance –Weather –Competition –Other.
1 Patent information for strategic technology management 作者: Holger Ernst 報告者:楊易霖 World Patent Information 25 (2003) 233–242.
Investigating the Accuracy and Robustness of the Icelandic Cod Assessment and Catch Control Rule A. Rosenberg, G. Kirkwood, M. Mangel, S. Hill and G. Parkes.
Monte Carlo Simulation and Personal Finance Jacob Foley.
Pacific Hake Management Strategy Evaluation Joint Technical Committee Northwest Fisheries Science Center, NOAA Pacific Biological Station, DFO School of.
GADGET - Globally applicable Area Disaggregated General Ecosystem Toolbox, Bjarte Bogstad, Institute of Marine Research, Bergen, Norway.
Slide 1 Estimating Performance Below the National Level Applying Simulation Methods to TIMSS Fourth Annual IES Research Conference Dan Sherman, Ph.D. American.
Statistics for Managers Using Microsoft Excel, 5e © 2008 Pearson Prentice-Hall, Inc.Chap 8-1 Statistics for Managers Using Microsoft® Excel 5th Edition.
Use of multiple selectivity patterns as a proxy for spatial structure Felipe Hurtado-Ferro 1, André E. Punt 1 & Kevin T. Hill 2 1 University of Washington,
Mean and Standard Deviation of Grouped Data Make a frequency table Compute the midpoint (x) for each class. Count the number of entries in each class (f).
VI. Evaluate Model Fit Basic questions that modelers must address are: How well does the model fit the data? Do changes to a model, such as reparameterization,
Evaluation of a practical method to estimate the variance parameter of random effects for time varying selectivity Hui-Hua Lee, Mark Maunder, Alexandre.
1 Multiple Regression Analysis y =  0 +  1 x 1 +  2 x  k x k + u.
. Assessment of the Icelandic cod stock Björn Ævarr Steinarsson Marine Research Institute.
Stock assessment of jack mackerel (Trachurus murphyi): a non- homogenous stock and changes in catchability. Hugo Arancibia* and Liesbeth van der Meer**
The Stock Synthesis Approach Based on many of the ideas proposed in Fournier and Archibald (1982), Methot developed a stock assessment approach and computer.
Regression-Based Linkage Analysis of General Pedigrees Pak Sham, Shaun Purcell, Stacey Cherny, Gonçalo Abecasis.
USING INDICATORS OF STOCK STATUS WHEN TRADITIONAL REFERENCE POINTS ARE NOT AVAILABLE: EVALUATION AND APPLICATION TO SKIPJACK TUNA IN THE EASTERN PACIFIC.
Fisheries 101: Modeling and assessments to achieve sustainability Training Module July 2013.
Flexible estimation of growth transition matrices: pdf parameters as non-linear functions of body length Richard McGarvey and John Feenstra CAPAM Workshop,
Term 4, 2006BIO656--Multilevel Models 1 PROJECTS ARE DUE By midnight, Friday, May 19 th Electronic submission only to Please.
1 Prof. Dr. Rainer Stachuletz Multiple Regression Analysis y =  0 +  1 x 1 +  2 x  k x k + u 1. Estimation.
M.S.M. Siddeeka*, J. Zhenga, A.E. Puntb, and D. Pengillya
Statistics for Managers Using Microsoft Excel, 5e © 2008 Pearson Prentice-Hall, Inc.Chap 8-1 Statistics for Managers Using Microsoft® Excel 5th Edition.
The effect of variable sampling efficiency on reliability of the observation error as a measure of uncertainty in abundance indices from scientific surveys.
Extending length-based models for data-limited fisheries into a state-space framework Merrill B. Rudd* and James T. Thorson *PhD Student, School of Aquatic.
Modeling biological-composition time series in integrated stock assessments: data weighting considerations and impact on estimates of stock status P. R.
Using distributions of likelihoods to diagnose parameter misspecification of integrated stock assessment models Jiangfeng Zhu * Shanghai Ocean University,
Measurements and Their Analysis. Introduction Note that in this chapter, we are talking about multiple measurements of the same quantity Numerical analysis.
CAN DIAGNOSTIC TESTS HELP IDENTIFY WHAT MODEL STRUCTURE IS MISSPECIFIED? Felipe Carvalho 1, Mark N. Maunder 2,3, Yi-Jay Chang 1, Kevin R. Piner 4, Andre.
Selectivity and two biomass measures in an age-based assessment of Antarctic krill Doug Kinzey, George Watters NOAA/NMFS/SWFSC/AERD CAPAM Workshop, March.
MSE Performance Metrics, Tentative Results and Summary Joint Technical Committee Northwest Fisheries Science Center, NOAA Pacific Biological Station, DFO.
Lecture 10 review Spatial sampling design –Systematic sampling is generally better than random sampling if the sampling universe has large-scale structure.
Data weighting and data conflicts in fishery stock assessments Chris Francis Wellington, New Zealand CAPAM workshop, “ Data conflict and weighting, likelihood.
NWFSC A short course on data weighting and process error in Stock Synthesis Allan Hicks CAPAM workshop October 19, 2015.
Chapter 9 Sampling Distributions 9.1 Sampling Distributions.
Population Dynamics and Stock Assessment of Red King Crab in Bristol Bay, Alaska Jie Zheng Alaska Department of Fish and Game Juneau, Alaska, USA.
Survey Data Conflicts and Bias and Temporal Variation of Model Parameters of St. Matthew Island Blue King Crab J. Zheng, D. Pengilly and V. A. Vanek ADF&G,
Is down weighting composition data adequate to deal with model misspecification or do we need to fix the model? Sheng-Ping Wang, Mark N. Maunder National.
Fish stock assessment Prof. Dr. Sahar Mehanna National Institute of Oceanography and Fisheries Fish population Dynamics Lab November,
Chapter 7 Confidence Interval Estimation
SAFS Quantitative Seminar
Presentation transcript:

Some Insights into Data Weighting in Integrated Stock Assessments André E. Punt 21 October 2015 Index-1 length-4

Background Johnson et al. PFMC Sablefish Assessment “Integrated” models potentially involve numerous data sources: Indices of abundance (CPUE, surveys) Length-composition data Age-composition data Discards Mean body weight Conditional age-at-length data Moreover, each data source may be available for more than one “fleet”

Objectives Outline alternative methods of weighting for: Length- and age-composition data Conditional age-at-length data Evaluate the performance of these methods given model mis- specification.

Methods for tuning length (and age) composition data-I Let be the observed proportion of animals in length-class L during year y, and be the model-predicted proportion of animals in length-class L during year y. Under the assumption that length samples are multinomial (as is the case in Synthesis, ASAP, etc), the weight assigned to the data is the “effective sample size”, : where the are the input effective sample sizes.

Methods for tuning length (and age) composition data-II McAllister-Ianelli: This method sets the effective sample size by comparing the residual variance with the variance expected under a multinomial distribution: To compute an overall effective sample size,, it is necessary to average over the. Two options are commonly used: McAllister-Ianelli-1: McAllister-Ianelli-2:

Methods for tuning length (and age) composition data-III But residuals for length-compositions are seldom uncorrelated between length- classes – enter “Francis weighting”. The idea behind Francis weighting is to base on the mean age or length, i.e.: where is the mid-point of length-class L.

Methods for tuning conditional age-at-length-I Conditional age-at-length (CAL) data are (essentially) age-length keys. These data provide information on year-class strength and growth. CAL data are matrices by year, which makes application of standard weighting schemes difficult.

Methods for tuning conditional age-at-length-II Let be the observed proportion of animals in length-class L during year y that are of age a, and be the model-predicted proportion of animals in length-class L during year y that are of age a. Under the assumption that age samples are multinomial conditional on length, the negative log-likelihood is: where the are the input effective sample sizes.

Methods for tuning conditional age-at-length-III The McAllister-Ianelli and Francis methods can be extended (naively) to handle conditional age-at-length data: McAllister-Ianelli: McAllister-Ianelli-1: McAllister-Ianelli-2: Francis-A:

Methods for tuning conditional age-at-length-IV The Francis-A can be criticised because it treats each row of an age-length key as being independent. This is unlikely to be true. The Francis weighting method for length (and age) data can be generalized to age-length keys (Francis-B) by applying the basic algorithm to the mean age of the age-length key, i.e.: where: is the fraction of animals during year y observed to be in length-class L.

SIMULATION STUDY EVALUATION consciouslyenlightened.com

Simulation Details-I Spatial structure: One zone OR Two zones with spatial variation in F Fleet structure: Non-trawl fleet Trawl fleet Data (by fleet and zone): CPUE series (all years; CV = 0.1) Length frequencies (all years; = 100) Age-at-length data (50% of year-fleet-zone combinations; = 500) Logistics: 100 simulations Single-area estimation method Performance measure: spawning biomass (summed over zones).

Tuning algorithms McAllister-Ianelli-1: Tune the residual variance for the CPUE data and use the McAllister-Ianelli-1 method for both length and CAL data. McAllister-Ianelli-2: As for McAllister-Ianelli-1 except use the McAllister- Ianelli-2 method for both length and CAL data. Francis / Francis-A: As for McAllister-Ianelli-1 except use Francis weighting for the length data and Francis-A weighting for the CAL data. Francis / Francis-B: As for McAllister-Ianelli-1 except use Francis weighting for the length data and Francis-B weighting for the CAL data. Each tuning algorithm (except Francis / Francis-A*) is applied five times

Results: One-zone operating model The estimation model is not mis-specified so the correct effective sample sizes are known. This allows some questions about the “in principle” performance of the methods (and tuning algorithms) to be explored. Does estimation performance depend on the initial weights? Yes – results not shown here Does estimation performance depend on the tuning algorithm? Yes – results not shown here Which method for calculating weights performs best?

1.McAllister-Ianelli-1 is biased for both length-frequency and conditional age- at-length data. 2.McAllister-Ianelli-2 performs best at calculating effective sample sizes for length data (Francis is unbiased, but imprecise). 3.McAllister-Ianelli-2 performs best at calculating effective samples for conditional age-at-length data (Francis-A and Francis-B are unbiased, but imprecise). The one-zone operating model

The two-zone operating model 1.The untuned method performs poorer than when tuning is applied (except for when McAllister-Ianelli- 1 is applied). 2.McAllister-Ianelli-1 leads to the poorest performance. 3.Francis / Francis-B leads to estimates with least bias for final spawning biomass (and final / initial spawning biomass), but not by much.

The two-zone operating model 1.With model-specification: 1.Francis leads to lower weights than McAllister-Ianelli-1 2.Francis-B leads to lower weights than Francis-A and McAllister-Ianelli-2. 2.Francis and Francis-B are imprecise (compared to McAllister- Ianelli-2 and Francis-A). 3.We don’t know the correct effective sample size for this case.

Overall conclusions General General –Avoid McAllister-Ianelli-1 (averaging of effective sample sizes). –McAllister-Ianelli-2 (harmonic mean) performs adequately over all cases (but was not optimal when there was model mis- specification). –Francis / Francis-B was the least biased tuning algorithm, but the estimates of effective sample size showed the highest between- simulation variation

Questions & Acknowledgements Chris Francis is thanked for discussions that led to the Francis-A and Francis-B methods. This work was partially supported by NOAA grant NA10OAR