http://www.nasa.gov/topics/earth/features/climate_partners.html
Tuesday, June 21, 2011
New NASA grant to study effect of Arctic climate change on bowhead whales
http://www.nasa.gov/topics/earth/features/climate_partners.html
Wednesday, March 9, 2011
MARSS 2.1 ready for testing
Download from the MARSS Dev Site
The big update of MARSS to allow B and Z estimation is finally ready for testing. It'll be awhile before its ready for CRAN since I need to work on a couple new case studies and fix bugs and stuff that comes up in testing. I added a mock up of a new case study on dynamic factor analysis (sensu Alain Zuur) since that is of special interest to fisheries folks. See the editlog on the dev site to see the recent changes
MARSS is an R package which fits multivariate state-space models to time series data. It is different than other state-space packages in that it does maximization via a constrained Expectation-Maximization (EM) algorithm (an extension of the EM algorithm developed by Shumway and Stoffer 2000). The EM algorithm is slower (much) than the BFGS algorithm used in other packages, but can be more stable (less likely to get hung up on numerical problems) for some types of state-space models also the EM algorithm used to produce good initial conditions for Bayesian or Nelder-Mead algorithms.
New in version 2.1
* B and Z estimation
* Dynamic factor analysis (sensu Harvey 1989 and Zuur et al 2003)
* Allows degenerate Q and R (variances=0)
* Fewer constraints on what type of Q and R matrices you can fit
MARSS 2.1 still does not allow covariates. Including that is next on my to-do list, but it will be awhile. MARSS 2.1 also does not allow temporally varying parameters. Again, it's easy enough to include but is not high on my to-do list.
The big update of MARSS to allow B and Z estimation is finally ready for testing. It'll be awhile before its ready for CRAN since I need to work on a couple new case studies and fix bugs and stuff that comes up in testing. I added a mock up of a new case study on dynamic factor analysis (sensu Alain Zuur) since that is of special interest to fisheries folks. See the editlog on the dev site to see the recent changes
MARSS is an R package which fits multivariate state-space models to time series data. It is different than other state-space packages in that it does maximization via a constrained Expectation-Maximization (EM) algorithm (an extension of the EM algorithm developed by Shumway and Stoffer 2000). The EM algorithm is slower (much) than the BFGS algorithm used in other packages, but can be more stable (less likely to get hung up on numerical problems) for some types of state-space models also the EM algorithm used to produce good initial conditions for Bayesian or Nelder-Mead algorithms.
New in version 2.1
* B and Z estimation
* Dynamic factor analysis (sensu Harvey 1989 and Zuur et al 2003)
* Allows degenerate Q and R (variances=0)
* Fewer constraints on what type of Q and R matrices you can fit
MARSS 2.1 still does not allow covariates. Including that is next on my to-do list, but it will be awhile. MARSS 2.1 also does not allow temporally varying parameters. Again, it's easy enough to include but is not high on my to-do list.
Wednesday, February 16, 2011
Winter session of the NWFSC Stats Reading group has started
This quarter's book is
Royle, J.A. and R.M. Dorazio. 2008. Hierarchical Modeling and Inference in Ecology: The Analysis of Data from Populations, Metapopulations, and Communities. Academic Press, San Diego, CA. xviii, 444 pp
Stephanie Hampton talking at ASLO 2011 about some joint work on plankton dynamics
CHALLENGES AND SOLUTIONS TO ANALYZING MARINE COMMUNITIES WITH MULTIVARIATE AUTOREGRESSIVE (MAR) MODELS
Multivariate autoregressive (MAR) models have a long history of use in economics, and have been used successfully to understand trophic interactions and community stability in freshwater plankton communities. Uses of MARs in marine ecosystems have been limited, however, possibly due to several challenges that arise with marine data. For example, many marine monitoring programs 1) use less consistent sampling locations, 2) have more variable time intervals, 3) may have higher observation error, and 4) include highly mobile or transient taxa. Comparison of results from standard MARs applied to point and transect data suggested that information loss in applying standard MARs to transect data can be substantial. Additionally, MAR results typically indicate more trophic links in freshwater communities than comparable marine communities, suggesting that modifications to MARs are necessary when analyzing typical marine data. Development of state-space MARs may help elucidate patterns in marine data by explicitly estimating species interactions in light of both process and observation errors. Extension of Bayesian methods also allows for probabilistic statements about parameters. We demonstrate these improvements in MARs via analyses of several plankton datasets.
This work associated with a CAMEO grant to S. Hampton, S. Katz, E. Holmes, and M. Scheuerell.
Multivariate autoregressive (MAR) models have a long history of use in economics, and have been used successfully to understand trophic interactions and community stability in freshwater plankton communities. Uses of MARs in marine ecosystems have been limited, however, possibly due to several challenges that arise with marine data. For example, many marine monitoring programs 1) use less consistent sampling locations, 2) have more variable time intervals, 3) may have higher observation error, and 4) include highly mobile or transient taxa. Comparison of results from standard MARs applied to point and transect data suggested that information loss in applying standard MARs to transect data can be substantial. Additionally, MAR results typically indicate more trophic links in freshwater communities than comparable marine communities, suggesting that modifications to MARs are necessary when analyzing typical marine data. Development of state-space MARs may help elucidate patterns in marine data by explicitly estimating species interactions in light of both process and observation errors. Extension of Bayesian methods also allows for probabilistic statements about parameters. We demonstrate these improvements in MARs via analyses of several plankton datasets.
This work associated with a CAMEO grant to S. Hampton, S. Katz, E. Holmes, and M. Scheuerell.
Tuesday, February 1, 2011
New isotope mixing model paper
Ward, E.J.*, Semmens, B.X., Phillips, D.L., and Moore, J.W. 2011. A quantitative approach
for grouping sources in stable isotope mixing models. In press, Ecosphere
* joint 1st authorship
Tuesday, October 19, 2010
MARSS 1.1 posted to CRAN
An updated version of MARSS which fixes a few bugs and adds some features has been posted to CRAN today: MARSS 1.1 Main issues fixed are:
- * Allow NA and NaN as miss.value
- * Fixed bug that prevented Monte Carlo initialization searches
- * Changed convergence test to use log param vs log iter (a standard test for actual convergence). Before we used new.LL-old.LL <= tol, but that is not a test for convergence actually. Now the fitting takes longer, but stops at convergence (or maxit).
- * Fixed bug that was preventing certain types of non-design Z matrices
- * Fixed summary(marssm). The labeling was off.
- * Added function for likelihood profiling and changing defaults
Monday, October 18, 2010
Brice talking this week on using multivariate time-series models to combine multiple data sources
Integrating time-series of community monitoring data
Semmens, BX, EE Holmes, EJ Ward, CV Pattengill-Semmens
Linking Science to Management, Oct 19-22, 2010, Duck Key, FL
http://conference.ifas.ufl.edu/floridakeys/index.html
Assessing population trends, evaluating management actions, and identifying community responses to anthropogenic impacts all require an accurate time-series of populations. In practice, such data are often scarce or of varying quality due to the limited resources of managing agencies. In such situations, analyses that integrating multiple data sources (e.g. agency monitoring programs, citizen science observations, fisheries catch records) can yield dramatic improvements in the estimation of population trajectories. To do so effectively, however, such integrative models must account for differences in observation errors across data sources. We used multivariate state space models (MSSMs) to assess the population trajectories of reef fish species from the Florida Keys National Marine Sanctuary based on data from 1) point count surveys conducted through academic institutions and 2) citizen-science monitoring surveys conducted by volunteer Scuba divers. By developing competing models and applying information theory, we demonstrate how MSSMs can be used to compare and integrate multiple monitoring time series, and ultimately improve estimates of the true states of populations though time. Additionally, we demonstrate that by combining multiple time series, it is possible to recover method-specific observation error estimates even for very short time series of data.
Assessing population trends, evaluating management actions, and identifying community responses to anthropogenic impacts all require an accurate time-series of populations. In practice, such data are often scarce or of varying quality due to the limited resources of managing agencies. In such situations, analyses that integrating multiple data sources (e.g. agency monitoring programs, citizen science observations, fisheries catch records) can yield dramatic improvements in the estimation of population trajectories. To do so effectively, however, such integrative models must account for differences in observation errors across data sources. We used multivariate state space models (MSSMs) to assess the population trajectories of reef fish species from the Florida Keys National Marine Sanctuary based on data from 1) point count surveys conducted through academic institutions and 2) citizen-science monitoring surveys conducted by volunteer Scuba divers. By developing competing models and applying information theory, we demonstrate how MSSMs can be used to compare and integrate multiple monitoring time series, and ultimately improve estimates of the true states of populations though time. Additionally, we demonstrate that by combining multiple time series, it is possible to recover method-specific observation error estimates even for very short time series of data.
Subscribe to:
Posts (Atom)