Final Technical Report: - Southwest Fisheries Science Center - NOAA

More documents

Recommendations

Info

4.2 Modeling Framework : GLM and GAM 4.2.1 Comparisons of GAM Algorithms During the comparison of GAM algorithms, we found a bug in the step.gam function from the R package gam code that previously had not been reported to the R mailing lists, and that was unknown to the package developer (pers comm. with Hastie). The bug prevented step.gam from including the offset term for survey effort in any encounter rate model that was examined during the stepwise search. As a result, we only modeled group size (and not encounter rates) using the step.gam algorithm from R package gam. The group size GAMs built using the S-PLUS and R package gam algorithms were essentially identical: the best models contained the exact same predictor variables and associated degrees of freedom, and the parameterization of the smoothing splines were identical, except for small differences that were likely due to the precision of the software platforms. GAMs built using R package mgcv were more variable. The mgcv gam algorithm allows users to adjust more parameters and settings to build the models compared to the S-PLUS analogue. To the knowledgeable user, this flexibility enables fine-tuning of the GAMs. On the other hand, having numerous adjustable arguments makes the algorithm less user-friendly because a greater investment of time must be spent to learn how to build appropriate models. Tables 10 and 11 show the range of encounter rate and group size models, respectively, selected as the final model by mgcv gam given the specified combination of settings for the gam.method, smoothing spline, and gamma arguments. The paired models for each species/response variable that are provided in these tables were chosen based on the sum of the absolute value of the deviation of the observed-to-predicted ratios of the response variable in the geographic strata shown in Figure 7. The “simple models” in Tables 10 and 11 represent the models having relatively few effective degrees of freedom and the smallest sum of absolute deviations of the observed-to-predicted ratios. Similarly, the “complex models” represent those having a relatively large number of effective degrees of freedom in addition to good agreement between observed and predicted values of the response variable. For cases in which a single model clearly outperformed all of the others, only one model is presented in the table. The variability in model complexity can be illustrated using the rough-toothed dolphin encounter rate models, where the preferred simple model had 8.9 degrees of freedom and the preferred complex model had over fifty degrees of freedom. The sum of absolute deviations of the observed-to-predicted ratios is smaller for the complex model. This is to be expected because the data used for predictions were also used to build the models; in this scenario, a complex model is more likely to exhibit fidelity to the data. 48
When cetacean experts were shown geographic contour plots of the predictions from the competing simple and complex mgcv gam models for each species during the SWFSC Cetacean Experts’ Workshop, the simple models were overwhelmingly preferred to the complex models. The dominant criticisms of the complex models from the expert panel were twofold: the predictions from the complex models either 1) exhibited relatively small-scale details in population density that are unexplainable given existing knowledge of the dynamics of the ecosystem, or 2) were nearly identical to those from the simple model and, therefore, the extra model complexity was not necessary for capturing the spatial patterns. Overall conclusions to be made from this investigation into the behavior of mgcv gam (summarized in Tables 10 and 11) are as follows: 4.2.2 Encounter Rate Models • The gam.method perf.magic produced the simple models with the greatest predictive performance. The best complex models were developed using outer (6 models), perf.outer (4 models), and perf.magic (2 models). • Cubic regression splines were preferred for building simple encounter rate models, whereas the complex models were constructed using either cubic or thin plate regression splines. • To our surprise, the preferred simple models were split almost equally between those built using gamma = 1.0 (8 models) and 1.4 (6 models). The best complex models were generally constructed using gamma = 1.0. • The sum of absolute deviations of the observed-to-predicted ratios was smaller for the complex models in most instances, although this is to be expected because the predictions were based on the same data used to build the models for this exercise. 4.2.3 Group Size Models • The gam.method magic produced the simple models with the greatest predictive performance. The best complex models were divided among gam.methods mgcv and magic. • The preferred simple models were constructed by thin plate regression splines, in general, whereas cubic regression splines were found in more of the preferred complex models. • The gamma parameter performed close to our expectations in the group size models, with the majority of simple models constructed using gamma = 1.4 and the majority of complex models using the default value of 1.0. • The trend in the sum of absolute deviations of the observed-to-predicted ratios was similar to that found for the encounter rate models, with simple models tending to have slightly larger values. 49
Page 1 and 2:
U N I R A P E D T T E M D S E N TA
Page 3:
C O L A N IO T A N U N C I EA .S. D
Page 6 and 7:
This report was prepared under cont
Page 8 and 9:
Table of Contents Acronyms and Abbr
Page 10 and 11:
C.1 Journal Publications ..........
Page 12 and 13:
Figure 18. Encounter rate models bu
Page 14 and 15:
List of Tables Table 1. Summary of
Page 16 and 17:
Appendix B, Table B-6. Summary of m
Page 18 and 19:
Acknowledgements This project was f
Page 20 and 21: of these choices as possible and us
Page 22 and 23: Although our models include most of
Page 24 and 25: 2.0 Background The Navy and other m
Page 26 and 27: comparison is based on a separate s
Page 28 and 29: Figure 1. Transects (green lines) s
Page 30 and 31: used to evaluate the ability of sum
Page 32 and 33: 3.1.5 Mid-trophic Sampling with Net
Page 34 and 35: variable. Interpolated maps of the
Page 36 and 37: Table 2. Variogram model results. A
Page 38 and 39: “number of individuals” as the
Page 40 and 41: Encounter Rate and Group Size Model
Page 42 and 43: were built separately for each of t
Page 44 and 45: Expanding models to the entire U.S.
Page 46 and 47: Table 4. Summary of the weighted ef
Page 48 and 49: Consistent with Barlow and Forney (
Page 50 and 51: incorporation of geographic coordin
Page 52 and 53: overwhelming majority of “Bryde
Page 54 and 55: the genera Ziphius and Mesoplodon.
Page 56 and 57: Table 7. Geographically stratified
Page 58 and 59: above, and the data themselves, are
Page 60 and 61: (hereafter “summer”). SWFSC has
Page 62 and 63: avoid these problems, we decided to
Page 64 and 65: Figure 9. Thermocline depth (m) obs
Page 66 and 67: Attempts to adjust search parameter
Page 68 and 69: CAMMS 1991 PODS 1993 ORCAWALE 1996
Page 72 and 73: 4.2.4 Conclusions Regardings Modeli
Page 74 and 75: Table 10 cont. Comparison of the si
Page 76 and 77: Table 11 cont. Comparison of the si
Page 78 and 79: Figure 15. The transect lines used
Page 80 and 81: A) Striped dolphin 10 km B) Short-b
Page 82 and 83: dependent. For example, the number
Page 84 and 85: Figure 19. Predicted average densit
Page 86 and 87: ecosystem, eastern spinner dolphins
Page 88 and 89: Table 13. Variables selected for mo
Page 90 and 91: Table 14. Starting and final AIC va
Page 92 and 93: Table 17. Ratios of observed to pre
Page 94 and 95: species represented in the acoustic
Page 96 and 97: 0.29 (Risso’s dolphin) to 3.20 (n
Page 98 and 99: Figure 25. Sample 2005 validation p
Page 100 and 101: Table 18. (continued) 1991 1993 Nor
Page 102 and 103: Blue whales had the greatest deviat
Page 104 and 105: Table 20. Abundance (number of anim
Page 106 and 107: Table 22. Proportion of deviance ex
Page 108 and 109: Figure 27. Average density (AveDens
Page 110 and 111: Table 23. Effective degrees of free
Page 112 and 113: We attempted to build encounter rat
Page 114 and 115: 5.0 Conclusion The field of predict
Page 116 and 117: 6.0 Transition Plan The models of c
Page 118 and 119: needed to ensure that the SDSS rema
Page 120 and 121:
Administrative Report LJ-07-02. NOA
Page 122 and 123:
Forney KA, Barlow J (1993) Prelimin
Page 124 and 125:
Redfern JV, Barlow J, Ballance LT,
Page 126 and 127:
Appendix A: Detailed Model Results
Page 128 and 129:
Table A-1 (continued) Blue whale 19
Page 130 and 131:
Figure A-1a. Striped dolphin 108
Page 132 and 133:
Figure A-1c. Risso’s dolphin 110
Page 134 and 135:
Figure A-1e. Northern right whale d
Page 136 and 137:
Figure A-1g. Sperm whale 114
Page 138 and 139:
Figure A-1i. Blue whale 116
Page 140 and 141:
Figure A-1k. Baird’s beaked whale
Page 142 and 143:
Figure A-2. Predicted average densi
Page 144 and 145:
Figure A-2b. Short-beaked common do
Page 146 and 147:
Figure A-2d. Pacific white-sided do
Page 148 and 149:
Figure A-2f. Dall’s porpoise 126
Page 150 and 151:
Figure A-2h. Fin whale 128
Page 152 and 153:
Figure A-2j. Humpback whale 130
Page 154 and 155:
Figure A-2l. Small beaked whales 13
Page 156 and 157:
Table B-2. Summary of model validat
Page 158 and 159:
Page 160 and 161:
Page 162 and 163:
Page 164 and 165:
Page 166 and 167:
Page 168 and 169:
Page 170 and 171:
Page 172 and 173:
Table B-10. Summary of model valida
Page 174 and 175:
Page 176 and 177:
Page 178 and 179:
Page 180 and 181:
Page 182 and 183:
Page 184 and 185:
Figure B-1. Predicted yearly and av
Page 186 and 187:
Figure B-1. b) Eastern spinner dolp
Page 188 and 189:
Figure B-1. c) Whitebelly spinner d
Page 190 and 191:
Figure B-1. d) Striped dolphin 168
Page 192 and 193:
Figure B-1. e) Rough-toothed dolphi
Page 194 and 195:
Figure B-1. f) Short-beaked common
Page 196 and 197:
Figure B-1. g) Bottlenose dolphin 1
Page 198 and 199:
Figure B-1. h) Risso’s dolphin 17
Page 200 and 201:
Figure B-1. i) Cuvier’s beaked wh
Page 202 and 203:
Figure B-1. j) Blue whale 180
Page 204 and 205:
Figure B-1. k) Bryde’s whale 182
Page 206 and 207:
Figure B-1. l) Short-finned pilot w
Page 208 and 209:
Figure B-1. m) Dwarf sperm whale 18
Page 210 and 211:
Figure B-1. n) Mesoplodon beaked wh
Page 212 and 213:
Figure B-1. o) Small beaked whales
Page 214 and 215:
Figure B-2. Predicted average densi
Page 216 and 217:
Figure B-2. (cont.) c) Whitebelly s
Page 218 and 219:
Figure B-2. (cont.) g) Bottlenose d
Page 220 and 221:
Figure B-2. (cont.) k) Bryde’s wh
Page 222 and 223:
Figure B-2. (cont.) o) Small beaked
Page 224 and 225:
Ferguson MC (2005) Cetacean Populat
Page 226 and 227:
Barlow J, Kahru M, Mitchell BG (In
Page 228 and 229:
Memorandum NMFS-SWFSC-374, U.S. Dep
show all

Final Technical Report: - Southwest Fisheries Science Center - NOAA

You also want an ePaper? Increase the reach of your titles

Delete template?

Save as template?