a synthesis of annual estimates of tir and d for wild populations presenter: paul wilson css annual...
DESCRIPTION
Synthesizing multi-study data In various fields, incl. ecology, information from multiple studies is combined using “meta-analysis” Goal is to increase power to detect effects, by estimating “summary effect” from individual effect size estimates Treat each migration year as a “study” and use meta-analysis techniquesTRANSCRIPT
A Synthesis of Annual Estimates of TIR and D for Wild Populations
Presenter: Paul Wilson CSS Annual Meeting Apr 2nd
2010
Chapter 5 ObjectivesMany years, CIs of TIRs & Ds wide
and/or contain 1—how can we determine if TIR or D is generally > or < 1?
Can we detect a in mean TIR or D due to altered management?
Inter-annual variation in TIR and D of interest for modeling viability & effects of management. Can we get a good estimate?
Synthesizing multi-study data
In various fields, incl. ecology, information from multiple studies is combined using “meta-analysis”
Goal is to increase power to detect effects, by estimating “summary effect” from individual effect size estimates
Treat each migration year as a “study” and use meta-analysis techniques
Kinds of meta-analysis“Fixed-effect”: all studies assumed to
share same true effect—variation in estimates due solely to experimental error
“Random-effects”: true effect size varies from study to study—summary effect estimates the mean of distribution of effect sizes
Annual TIR & D values surely vary, so we use random effects formulas
Methods
RE Meta-analysis on TIR & D
TIR and D = “response ratios”. Use ln(TIR) & ln(D) to linearize & normalize
Need to estimate “between-study” variance: variance in true effect size between years
Individual years weighted by inverse of (annual sampling variance + between-year variance)
Estimate mean, CI of mean, & variance of distributions of TIR and D for years included
Data for analysisData for wild Chinook 1994-2006 M.Y.sWild steelhead: 1997-2006 M.Y.sBoth species, did two analyses for TIR
& D: 1) including 2001 M.Y., 2) omitting 2001
For Chinook TIR only, also did analysis using only C0 fish for in-river SAR
Used bootstrap output for annual sampling variance, except for C0-only analysis
Results
Forest plot of Chinook TIR, incl. 2001 (C1)
Year1994199519961997199819992000200120022003200420052006
Summary
TIR1.620.951.920.740.871.140.608.960.651.061.092.140.79___ 1.17
0.5 1.5 2.5 3.5 4.5 6.5 8.5 11.5
TIR
Forest plot of Chinook TIR, without 2001
Year199419951996199719981999200020022003200420052006
Summary
TIR1.620.951.920.740.871.140.600.651.061.092.140.79___ 0.99
0.5 1.5 2.5 3.5 4.5 5.5 7.5 9.5 12.5
TIR
Forest plot of C0-only Chinook TIR
Year1994199519961997199819992000200120022003200420052006
Summary
TIR1.620.951.920.740.871.140.602.960.651.061.093.160.79____0.94
0.5 1.5 2.5 3.5 4.5 6.5 8.5 11.5
TIR
Forest plot of steelhead TIR, incl. 2001 (C1)
Year1997199819992000200120022003200420052006
Summary
TIR2.200.202.281.4537 4.254.4114.34.880.98___ 3.26
0.1 1.0 2.0 3.0 5.0 8.012.0 19.029.044.0
TIR
Forest plot of steelhead TIR, without 2001
Year199719981999200020022003200420052006
Summary
TIR2.200.202.281.454.254.4114.34.880.98___ 2.61
0.1 1.0 2.0 3.0 5.0 7.0 11.0 17.0 26.039.0
TIR
Summary mean TIR (center lines), 90% CIs of summary mean (boxes), and 90% prediction limits of summary TIR (whiskers).
0.1
1.0
10.0
100.0
Ck w/ 2001 Ck w/o 2001 Ck C0 only SH w/ 2001 SH w/o 2001
TIR
TIR findings summary 2001 highly influences Chinook B-Y var, if C1 fish
are used as in-river group that year Mean Ck TIR ~ sensitive to use of C1 fish and
inclusion of 2001, but all C.I.s include 1 SH B-Y var much > than Ck B-Y var; ~ sens. to
2001 Mean SH TIR >> 1 & C.I.s > 1, benefit from
transporting
Forest Plot of wild Chinook D, incl. 2001 (C1)
Year1994199519961997199819992000200120022003200420052006
Summary
D0.36 0.42 0.92 0.40 0.55 0.72 0.32 2.16 0.44 0.68 0.45 1.07 0.48 ___ 0.593
0.2 1.2 2.2 3.2 4.2
D
Forest Plot of wild Chinook D, without 2001
Year199419951996199719981999200020022003200420052006
Summary
D0.36 0.42 0.92 0.40 0.55 0.72 0.32 0.44 0.68 0.45 1.07 0.48 ___ 0.552
0.2 1.2 2.2 3.2 4.2
D
Forest Plot of wild steelhead D, incl. 2001 (C1)
Year1997199819992000200120022003200420052006
Summary
D1.180.111.070.501.462.241.752.691.300.60___ 1.12
0.02 1.00 2.00 3.00 5.00
D
Forest Plot of wild steelhead D, without 2001
Year199719981999200020022003200420052006
Summary
D1.180.111.070.502.241.752.691.300.60___ 1.09
0.02 1.00 2.00 3.00 5.00
D
Summary mean D (center lines), 90% CIs of summary mean (boxes), and 90% prediction limits of summary D (whiskers).
0.1
1.0
10.0
Ck w/ 2001 Ck w/o 2001 SH w/ 2001 SH w/o 2001
D
D findings summary 2001 strongly influences Chinook B-Y var Mean Ck D slightly sensitive to inclusion of
2001 Mean Ck D << 1, and both C.I.s < 1
delayed transportation mortality SH B-Y var much > than Ck B-Y var; not
sens. to 2001 SH mean D not sens. to 2001; C.I.s include
1
ConclusionsBetter estimates of mean than
unweighted Strong evidence that wild steelhead TIR
> 1, under operations prior to 2007, but benefit is highly variable. SH TIR > Ck TIR both because SH D > Ck D and SH SR < Ck SR.
Maximum transport strategies may not have maximized wild Chinook SAR
Delayed transport mortality of wild Chinook
Unclear if wild steelhead experience delayed transport mortality
Caveats & ExtensionsSARs of adjacent year classes not
independent-but only ratios of SARs analyzed
Dependence means CIs too narrow. Likely doesn’t change main conclusions
Marking more steelhead marking could help narrow CIs
Could use “meta-regression” to further investigate relation of TIR to SR
Questions?