Aster Models for Life History Analysis

University of Minnesota, Twin Cities     School of Statistics     Charlie Geyer's Home Page
Department of Ecology, Evolution, and Behavior     Ruth Shaw's Home Page
Chicago Botanic Garden     Institute for Plant Conservation Biology     Stuart Wagenius's Home Page
Echinacea Project Page

Last changed: Wed May 5 19:04 CDT 2010

Contents

News

New! Version 0.8-27 of the aster package is on CRAN. This version provides a new function anova.asterOrReaster that does likelihood ratio tests of model comparison of fits done by the aster and reaster functions. It also now checks that models are nested (as they must be for such tests to make sense).

New! Version 0.8-23 of the aster package is on CRAN. This version fixes a bug that the summary.reaster function gave an inexplicable error message when all variance component estimates were zero.

New! Version 0.8-20 of the aster package has shipped to CRAN. This version fixes an issue raised by John Stanton-Geddes. Having NA values in aster data is almost always a mistake. So this version make it an error. The help pages for the aster function and the reaster function have a new section about this.

New! Version 0.8-19 of the aster package is on CRAN now. This version does random effects! For more info see under The R Package and the new Technical Report 692.

The second version of the aster package, which is called aster2 has has shipped! This is still a beta version and no one should use it yet unless they need models with dependence groups, which aster2 implements whereas aster does not. The new package is not backwards compatible with the old but should be a lot easier to use. Thus when the new package is feature complete (not yet, but real soon now) it should be used for all new projects. The old package will not have any new features added but will get bug fixes, if required, and will remain on CRAN so old analyses can be redone.

We have started putting technical reports on the University of Minnesota Digital Conservancy. We do not expect this site to go away anytime soon, but the conservancy will be there permanently.

The paper Unifying Life History Analyses for Inference of Fitness and Population Growth has been given the 2009 Presidential Award of the American Society of Naturalists. This award is for the best paper published in The American Naturalist during the calendar year preceding the President's term of office. The President of the American Society of Naturalists (ASN), in this year Joel Kingsolver, made this award.

R Package

The software is now at CRAN. The package name is aster. Now at version 0.8-27.

To install the software in an installation of R which you own (not the Administrator or root user), just do

install.packages("aster")

at an R command line. Otherwise, ask your friendly local system administrator to do this.

Since version 0.3 the package includes an example dataset and examples on all the help pages for functions we expect users to use. Since version 0.7-2 the package includes four more example datasets echin2 and chamae and chamae2 and aphid, which are used in a new technical report. Since version 0.7-7 the package includes one more example dataset sim, which was simulated in technical report 669 and used again in technical report 674. Since version 0.8-19 the package includes a new function reaster, which fits aster models with random effects using approximate maximum likelihood, and several new datasets (used for random effects examples), chamae3 and oats and radish.

The library also contains a hopefully useful tutorial (package vignette in R terminology).

Papers, Tech Reports, and Tutorial

The paper

Charles J. Geyer, Caroline E. Ridley, Robert G. Latta, Julie R. Etterson, and Ruth G. Shaw (2013).
Local Adaptation and Genetic Effects on Fitness: Calculations for Exponential Family Models with Random Effects.
Annals of Applied Statistics, 7, 1778–1795.

which was formerly titled Aster Models with Random Effects, is now available at the Annals of Applied Statistics web site. The technical report (692) cited therein is linked below.

The paper

Ruth G. Shaw and Charles J. Geyer (2010).
Inferring fitness landscapes.
Evolution 64, 2510–2520.

is now available at the Evolution web site. The five technical reports (669, 670, 671, 674, and 675) cited therein are linked below.

The paper

Stuart Wagenius, Helen H. Hangelbroek, Caroline E. Ridley, and Ruth G. Shaw (2010).
Biparental inbreeding and inter-remnant mating in a perennial prairie plant: fitness consequences for progeny in their first eight years.
Evolution, 64, 761–771.

is now available at the Evolution web site.

The paper

Ruth G. Shaw, Charles J. Geyer, Stuart Wagenius, Helen H. Hangelbroek, and Julie R. Etterson (2008).
Unifying Life History Analyses for Inference of Fitness and Population Growth.
American Naturalist, 172, E35–E47.

is now available, as are three technical reports that contain the full data analysis for the examples in the paper.

The paper

Charles J. Geyer, Stuart Wagenius, and Ruth G. Shaw (2007).
Aster Models for Life History Analysis.
Biometrika, 94, 415–426.

is now available, as is a technical report that contains the full data analysis for the example in the paper and many theoretical details left out of the paper, as is a tutorial (package vignette) that gives a more basic introduction to the software.

PDF File Description
TR 696 New! Technical report titled Aster Models with Random Effects and Additive Genetic Variance for Fitness, which shows how to calculate additive genetic variance for fitness, which is needed to apply Fisher's fundamental theorem of natural selection. (Two files are at the University of Minnesota Digital Conservancy, a PDF of the tecnical report and the Rnw file needed to recreate the TR, which includes all R code for all calculations.)
TR 692 New! Technical report titled Aster Models with Random Effects via Penalized Likelihood, which introduces aster models with (approximate) random effects. (Two files are at the University of Minnesota Digital Conservancy, a PDF of the tecnical report and the Rnw file needed to recreate the TR, which includes all R code for all calculations.)
TR 676 Technical report titled A Philosophical Look at Aster Models, which gives an overview of the mathematical statistics behind aster models without too much technicality. (One file at the University of Minnesota Digital Conservancy, a PDF of the tecnical report).
TR 675 revised Technical report titled Aster Models and Lande-Arnold Beta, which gives valid hypothesis and confidence intervals for directional selection gradients. (Two files are at the University of Minnesota Digital Conservancy, a PDF of the tecnical report and the Rnw file needed to recreate the TR, which include all R code for all calculations.) This technical report was revised January 13, 2010. If you have the earlier version, be advised that figure is wrong. (Just for the record, the earlier version can be found here but should not be used for any reason other than comparing to see what the error was.)
TR 674 revised Technical report titled Hypothesis Tests and Confidence Intervals Involving Fitness Landscapes fit by Aster Models, of how to do hypothesis tests about whether the fitness landscape has a maximum (stabilizing selection exists) and, if so, how to do confidence regions for the location of the maximum. (Two files are at the University of Minnesota Digital Conservancy, a PDF of the tecnical report and the Rnw file needed to recreate the TR, which include all R code for all calculations.) This technical report was revised January 9, 2010. If you have the earlier version, be advised that confidence region for the maximum is too large (square root forgotten). (Just for the record, the earlier version can be found here but should not be used for any reason other than comparing to see what the error was.)
TR 671 revised Technical report titled Model Selection in Estimation of Fitness Landscapes, which is a detailed explication of what to do when many phenotypic traits are measured and, although one wants model fitness as a function of all of them, to get good statistical estimation one must use submodels with fewer parameters. (Two files are at the University of Minnesota Digital Conservancy, a PDF of the tecnical report and the Rnw file needed to recreate the TR, which include all R code for all calculations.) This technical report was revised July 6, 2009. If you have the earlier version, be advised that the frequentist model averaging calculations in the earlier version are wrong. (Just for the record, the earlier version can be found here but should not be used for any reason other than comparing to see what the error was.)
TR 670 Technical report titled Commentary on Lande-Arnold Analysis, which is a detailed explication of the theory in Lande and Arnold (1983) and also has some supplementary plots for a talk given at Evolution 2008, University of Minnesota, June 20–24. (Two files are at the University of Minnesota Digital Conservancy, a PDF of the tecnical report and the Rnw file needed to recreate the TR, which include all R code for all calculations.)
TR 669 Technical report having all the calculations and plots for a talk given at Evolution 2008, University of Minnesota, June 20–24. (Two files are at the University of Minnesota Digital Conservancy, a PDF of the tecnical report and the Rnw file needed to recreate the TR, which include all R code for all calculations.)
aster2.rgs.pdf The second paper on aster analysis. This one contains a review of life history analysis and how aster analysis fits in and three examples, including analysis of fitness landscapes and estimation of population growth rate. This has been revised and resubmitted to American Naturalist. (Of historical interest only, the first submitted version)
tr666.pdf Technical report that has yet another calculation for one example for the on-line supplement in the second (and final) resubmission of the paper by Shaw, Geyer, Wagenius, Hangelbroek, and Etterson described above. Files needed to recreate the TR, which include all R code for all calculations, are in this folder.
tr661.pdf Technical report that has the the calculations for one example in the revised resubmission of the paper by Shaw, Geyer, Wagenius, Hangelbroek, and Etterson described above. (This TR was slightly revised, Nov 24, 2007, revised again, adding one more page and plot at the end, Nov 26, 2007, then revised yet again Dec 26, 2007 changing only some plot labels.) Files needed to recreate the TR, which include all R code for all calculations, are in this folder.
tr658.pdf Technical report that has all of the calculations for the rest of the examples in the revised resubmission of the paper by Shaw, Geyer, Wagenius, Hangelbroek, and Etterson described above (plus calculations for the other example in the original submission, which are improved by the calculations in TR 661). Files needed to recreate the TR, which include all R code for all calculations, are in this folder.
aster-submit3.pdf revised and resubmitted paper, double spaced. This paper has appeared in Biometrika. The printed version is available from Oxford University Press. (Of historical interest only, the first submitted version and the second submitted version)
tr644.pdf Technical report that has the the calculations for one example in the revised resubmission of the paper by Geyer, Wagenius, and Shaw described above.
tutor.pdf tutorial (package vignette)
white.pdf expansion of last section of tutorial about parametric bootstrap
multi.pdf technical report (653) allowing correlated child nodes. Of historical interest only, incorporated in the published version of the paper by Geyer, Wagenius, and Shaw described above.

Talks

Charlie Geyer gave a talk at the 2013 Midwest Statistics Research Colloquium in Madison, WI. Here are the slides for this talk.

Charlie Geyer gave a talk at the Department of Statistics, Rice University. Here are the slides for this talk.

Charlie Geyer gave a talk at the Seventh Annual Conference on Frontiers in Applied and Computational Mathematics (FACM '10) at the New Jersey Institute of Technology (NJIT) in Newark, New Jersey. Here are the slides for this talk.

Ruth Shaw gave a talk at Evolution 2008. Here are the slides for this talk.

Stuart Wagenius gave a talk at Evolution 2008. Here are the slides for this talk.

Charlie Geyer gave a talk at Evolution 2008. Here are the slides for this talk.

Charlie Geyer gave a talk at WNAR. Here are the slides for this talk.

Ruth Shaw gave a talk at the Peter Yodzis Colloquium. Here are the slides for this talk.

Stuart Wagenius gave a talk at the annual Ecology/ESA meeting August 2006, Memphis, TN titled Joint analysis of survival and reproduction over 10 years in perennial Echinacea plants from seven populations Here are the slides for this talk.

Charlie Geyer gave a talk in the School of Statistics seminar on March 30, 2006. Here are the slides for this talk.

Ruth Shaw gave a talk about this at the Society for the Study of Evolution meeting in Fairbanks, Alaska (June, 2005). Here are the original Microsoft PowerPoint form of the presentation and the Adobe PDF (Portable Document Format) form produced by OpenOffice.org.

Here is the full data analysis that provides the example for this talk. This has now been replaced by Appendix D of the aster modelling technical report.

The Name of the Game

The reason why I chose to call the package aster is that the organism for the data we are using for the initial analysis is the purple coneflower (Echinacea angustifolia) which is in the aster family. The submitted draft of the paper says

We named our models after flowers because the name is short and much nicer than forest graph exponential family conditional or unconditional canonical statistic models or any other descriptive name we could think of.

Not a very good reason perhaps, but bootstrap and jackknife and simulated annealing have even less connection to the actual statistics.

You can call it what you like, LHA for life history analysis or FEF for flat exponential family or whatever. But it would be too annoying at this point to change the name of the R package, since the function names

aster
summary.aster
anova.aster
predict.aster

should not change.

UROP Opportunites

UROP (Undergraduate Research Opportunities Program) opportunities involving aster models are described on this web page.

Valid HTML 4.01 Strict