lme4 convergence

From Bolker’s glmmFAQ

https://bbolker.github.io/mixedmodels-misc/glmmFAQ.html#convergence-warnings

 

A good walkthrough

https://rstudio-pubs-static.s3.amazonaws.com/33653_57fc7b8e5d484c909b615d8633c01d51.html

 

 

From stackexchange

http://stats.stackexchange.com/questions/110004/how-scared-should-we-be-about-convergence-warnings-in-lme4

 

http://stats.stackexchange.com/questions/164457/r-glmer-warnings-model-fails-to-converge-model-is-nearly-unidentifiable

 

http://stats.stackexchange.com/questions/97834/warning-messages-from-mixed-model-glmer

 

http://stackoverflow.com/questions/23478792/warning-messages-when-trying-to-run-glmer-in-r

 

https://stat.ethz.ch/pipermail/r-sig-mixed-models/2014q2/022026.html

Multivariate stats resources

 

Websites / Blogs

1. Ordination Methods for Ecologists

2. SAMPLE(ECOLOGY): NMDS TUTORIAL IN R (blog)

3. GUSTA ME: GUide to STatistical Analysis in Microbial Ecology

 

Videos

1. Methods in Ecology & Evolution blog: boral by Francis Hui

 

https://methodsblog.wordpress.com/2014/11/03/boral-r-package-for-multivariate-data-analysis-in-ecology/

The Morass of Multiple Comparisons

 

Minitab: “Which multiple comparison method should I use?”

The

Resources for multivariate statistics

LabDSV: R Labs for Vegetation Ecologists

• Ordination
  • Lab 7 Principal Components Analysis
  • Lab 8 Principal Coordinates Analysis
  • Lab 9 Nonmetric Multi-Dimensional Scaling
  • Lab 10 Correspondence Analysis and Detrended Corresponence Analysis
  • Lab 11 Fuzzy Set Ordination
  • Lab 12 Canonical Correspondence Analysis
• Cluster Analysis
  • Lab 13 Cluster Analysis
  • Lab 14 Discriminant Analysis with Tree Classifiers

Blogs:

PCa and PCoA explained

PCa and PCoA explained

Posted on January 17, 2012 by Bob O’H

GUSTA ME

http://occamstypewriter.org/boboh/2012/01/17/pca_and_pcoa_explained/

• • Lab 12 Canonical Correspondence Analysis
• Cluster Analysis
  • Lab 13 Cluster Analysis
  • Lab 14 Discriminant Analysis with Tree Classifiers

Interpreting Main Effecs & Interactions in ANOVAs

Below are some blogs related to the question of interactions in ANOVA.  They might not be super-accessible for ‘yins – they are mostly discussions among well practiced stats users and not attempts to explain the issue in general terms.  One thing I think is apparent though is that I think the problem is not well defined – when someone says “how do you interpret interactions” different people think of it in terms of experiments analyzed with an ANOVA with only categorical variables, while others think in terms of regression with interactions between two continuous variables, or between a continuous and categorical variable.  I also think some people think in terms of how you interpret the result of an ANOVA table (which basically just tells you p-values) and interpretation of the means and other parameters you can get from your model.

 

Dynamic Ecology: Interpreting ANOVA interactions and model selection: a summary of current practices and some recommendations (also see the comments on their Poll: How do you interpret two-way ANOVAs?)

Andrew Gelman: Main effects and interactions

PCA with categorical variables: Multiple correspondence analysis

5 functions to do Multiple Correspondence Analysis in R

This mentions the ade4 package, and factominer

 

The factorminer package also does mca

FactoMineR: Multiple Correspondence Analysis

 

 

Multiple Correspondence Analysis Essentials: Interpretation and application to investigate the associations between categories of multiple qualitative variables – R software and data mining

 

 

I think this is the reference that might be the origin of PCA w/ categorical variables in ecology related fields

 

Hill & Smith 1976 Principal component analysis of taxonomic data with multistate discrete characters

 

The package ade4 has a function dudi.mca() for multiple correspondence analysis (PCA w/categorical variables) and dudi.hillsmith() which allows you to do a mix; its probably similar to that PCAmix package I sent.  ade4 is part of a suite of packages by crazy French ecologists that are REALLY into multivariate stuff- I find most of the documentation really hard to understand (lots of math, not much ecology, not perfect English).  They have a paper “The ade4 package: implementing the duality diagram for ecologists” In the Journal for Statistical Software which has a section “4. An example: dudi.hillsmith” where they analyze the “dune meadow” dataset that shows up often in vegan I think.  This example, whether analyzed with dudi.hillmsith or the functions in the PCAmix package (or whatever its name is) might be a good place to start.

Simpson’s paradox

NWO, GENDER BIAS AND SIMPSON’S PARADOX

http://blog.casperalbers.nl/science/nwo-gender-bias-and-simpsons-paradox/

 

 

ONE MORE TIME: NWO, GENDER BIAS AND SIMPSON’S PARADOX

http://blog.casperalbers.nl/science/one-more-time-nwo-gender-bias-and-simpsons-paradox/

 

Dutch sexism study comes under files

http://news.vanquishmerchantbank.com/dutch-sexism-study-comes-under-fire/

 

 

 

Simpson’s paradox

https://en.wikipedia.org/wiki/Simpson%27s_paradox#Examples

 

Simpson’s paradox visualization

http://emilkirkegaard.dk/understanding_statistics/?app=Simpson_paradox

 

Simulating Data in R

R packages

simstudy

https://cran.r-project.org/web/packages/simstudy/vignettes/simstudy.html

psych::sim.multilevel

http://personality-project.org/r/html/sim.multilevel.html

 

Re-creating data sets

https://stats.stackexchange.com/questions/30303/how-to-simulate-data-that-satisfy-specific-constraints-such-as-having-specific-m

 

A complex but thorough simulation example from idre.ucla

http://stats.idre.ucla.edu/r/codefragments/mesimulation/

 

Mark-recapture designs for wildlife studies

https://sites.google.com/site/wild8390/software/simulate

 

Multilevel Simulation with specified ICC

https://psychometroscar.wordpress.com/simulate-a-2-level-multilevelhlmlinear-mixed-model/

 

Other

http://clayford.github.io/dwir/dwr_12_generating_data.html

https://web.stanford.edu/class/bios221/labs/simulation/Lab_3_simulation.html

http://stackoverflow.com/questions/23256694/how-can-i-efficiently-generate-a-dataframe-of-simulated-values

 

Click to access sim.pdf

 

http://www.quantumforest.com/2011/10/simulating-data-following-a-given-covariance-structure/

R2 in Generalized Linear mixed Models

Bolkers GLMM FAQ: Model summaries (goodness-of-fit, decomposition of variance, etc.) http://bbolker.github.io/mixedmodels-misc/glmmFAQ.html#how-do-i-compute-a-coefficient-of-determination-r2-or-an-analogue-for-glmms

Listserve thread: http://thread.gmane.org/gmane.comp.lang.r.lme4.devel/3281, including a comment by Doug Bates: http://thread.gmane.org/gmane.comp.lang.r.lme4.devel/3281

Listserve thread: http://thread.gmane.org/gmane.comp.lang.r.lme4.devel/684

Cross Validated

A problem that interests me: small marging $R^2$ values for otherwise interesting models:  Is it worth reporting small fixed-effect R2 (marginal R2), large model R2 (conditional R2)?

Partitioning explained variance to fixed effects by comparing r squared (R2) between linear mixed models

Is R2 useful or dangerous? (general issues with R2)

Misc. references on R2 in GLMMs at Proportion of explained variance in a mixed-effects model

Good explanation of Nakagawa and Schielzeth (2013) at: R2 for mixed models with multiple fixed and random effects

An open question: R2 for negative binomial GLMM: R2 from a generalized linear mixed-effects models (GLMM) using a negative binomial distribution

In response to the question Calculating R2 in mixed models using Nakagawa & Schielzeth’s (2013) R2glmm method, some re-posts a response from Douglas Bates where he voices his extreme skepticism about R2 for mixed models.

A related topics: Does the variance of a sum equal the sum of the variances?

R Packages for R2

muMIn::r.squaredGLMM

piecewiseSEM::rsquared

sjstats::r2

sjstats:cod “coefficient of discrimination” for logistic regression.  See Tjur T (2009) Coefficients of determination in logistic regression models – a new proposal: The coefficient of discrimination. The American Statistician, 63(4): 366-372

sjstats:rsme “root mean square error”

Documentation for sjstats discusses how ICC can be used to investigate amount of variance due to clustering

R Packages for related stuff

rptR: Repeatability estimation for Gaussian and non-Gaussian data; An introduction to repeatability estimation with rptR

References

Jaeger et al 2017.  An R² statistic for fixed effects in the generalized linear mixed model.  http://www.tandfonline.com/doi/abs/10.1080/02664763.2016.1193725?journalCode=cjas20

LaHuis et al.  2014.  Explained Variance Measures for Multilevel Models.  http://journals.sagepub.com/doi/abs/10.1177/1094428114541701

Tjur T (2009) Coefficients of determination in logistic regression models – a new proposal: The coefficient of discrimination. The American Statistician, 63(4): 366-372

http://www.tandfonline.com/doi/abs/10.1198/tast.2009.08210

 

Assessing the Fit of Regression Models

Measures of Model Fit for Linear Regression Models

lmer tools

merTools

https://cran.r-project.org/web/packages/merTools/vignettes/merToolsIntro.html

 

predictmeans: “This package provides functions to diagnose and make inferences from various linear models, such as … ‘lme’, and ‘lmer’. Inferences include predicted means and standard errors, contrasts, multiple comparisons, permutation tests and graphs.”

https://cran.r-project.org/web/packages/predictmeans/index.html

 

iccbeta: “This package quantifies the share of variance in a dependent variable that is attributed to group heterogeneity in slopes.”

Click to access iccbeta.pdf

 

influence.ME

https://cran.r-project.org/web/packages/influence.ME/index.html

 

lmerTest

https://cran.r-project.org/web/packages/lmerTest/index.html

 

longpower: sample size calculation for longitudinal data

https://cran.r-project.org/web/packages/longpower/index.html

 

mlmRev: example of lmer

https://cran.r-project.org/web/packages/mlmRev/index.html