similar to: variance explained by each term in a GAM

Hello fellow R's, I do apologize if this is a basic question. I'm doing some GAMs using the mgcv package, and I am wondering what is the most appropriate way to determine how much of the variability in the dependent variable is explained by each term in the model. The information provided by summary.gam() relates to the significance of each term (F, p-value) and to the

Hi, I am using mgcv:gam and have developed a model with 5 smoothed predictors and one factor. gam1 <- gam(log.sp~ s(Spr.precip,bs="ts") + s(Win.precip,bs="ts") + s( Spr.Tmin,bs="ts") + s(P.sum.Tmin,bs="ts") + s( Win.Tmax,bs="ts") +factor(site),data=dat3) The total deviance explained = 70.4%. I would like to extract the variance explained

Dear list, I have read several posts on this topic. I would use the same methodology as proposed by Simon Wood in this post: http://r.789695.n4.nabble.com/variance-explained-by-each-term-in-a-GAM-td836513.html My first question is: Does anyone know a scientific source (paper, book,...) that explains or uses this methodology. I have read several articles, particularly in the field of ecology,

Dear R users, I read your methods of extracting the variance explained by each predictor in different places. My question is: using the method you suggested, the sum of the deviance explained by all terms is not equal to the deviance explained by the full model. Could you tell me what caused such problem? > set.seed(0) > n<-400 > x1 <- runif(n, 0, 1) > ## to see problem

Dear list, I?m using the ?mgcv? package to fit some GAMs. Some of my covariates are derived quantities and have an associated standard error, so I would like to incorporate this uncertainty into the GAM estimation process. Ideally, during the estimation process less importance would be given to observations whose covariates have high standard errors. The gam() function in the ?mgcv? package

Dear list, I´m using the mgcv package to model the proportion by weight of certain prey on the stomach content of a predator. This proportion is the ratio of two weights (prey weight over stomach weight), and ranges between 0 and 1. The variance is low when proportion is close to 0 and 1, and higher at intermediate values. It seems that the best way to go is to model this using the

Hi all, This is pretty basic but I am not an expert and I couldn't find anything in the forum or my statistics book about it. I was reading a paper and the authors were using both "explained deviance" and "explained variance" as synonyms. They were describing a GAM regression. Is that right? I performed an analysis in R to take a look to the output of GAM regression and I

In Dr. Wood's book on GAM, he suggests in section 4.1.6 that it might be useful to shrink a single smooth by adding S=S+epsilon*I to the penalty matrix S. The context was the need to be able to shrink the term to zero if appropriate. I'd like to do this in order to shrink the coefficients towards zero (irrespective of the penalty for "wiggliness") - but not necessarily all the

I fit a GAM to turtle growth data using mgcv: >m1 <- gam(growth~s(mean.size, bs="cr")+s(year,bs="cr",k=7)+s(cohort,bs="cr")+s(age,bs="cr"), data=grow, family=quasi(link="identity")) The errors are skewed (and seem to be correlated with age) (code and plots here:

Dear list, I do apologize if these are basic questions. I am fitting some GAM models using the mgcv package and following the model selection criteria proposed by Wood and Augustin (2002, Ecol. Model. 157, p. 157-177). One criterion to decide if a term should be dropped from a model is if the estimated degrees of freedom (EDF) for the term are close to their lower limit. What would be the

Dear list, I?m wondering if there is something analogous to the TukeyHSD function that could be used for parametric terms in a GAM. I?m using the mgcv package to fit models that have some continuous predictors (modeled as smooth terms) and a single categorical predictor. I would like to do post hoc test on the categorical predictor in the models where it is significant. Any suggestions?

Hi, Does anyone know how to fit a GAM where one or more smooth terms are constrained to be monotonic, in the presence of "by" variables or other terms? I looked at the example in ?pcls but so far have not been able to adapt it to the case where there is more than one predictor. For example, require(mgcv) set.seed(0) n<-100 # Generate data from a monotonic truth.

Dear List, I've just tried to specify a GAM without an intercept -- I've got one of the (rare) cases where it is appropriate for E(y) -> 0 as X ->0. Naively running a GAM with the "-1" appended to the formula and the calling "predict.gam", I see that the model isn't behaving as expected. I don't understand why this would be. Google turns up this old

I am trying to compare and contrast the smoothing in the {mgcv} version of gam vs. the {gam} version of gam but I get a strange side effects when I try to alternate calls to these routines, even though I detach and unload namespaces. Specifically when I start up R the following code runs successfully until the last line i.e. plot(g4,se=TRUE) when I get "Error in dim(data) <- dim :

Hello, I am using the function step.gam() from the 'gam' package (header info from library(help=gam) included below) and have come across some behavior that I cannot understand. In short, I have written a function that 1) creates a dataframe, 2) calls gam() to create a gam object, then 3) calls step.gam() to run stepwise selection on the output from gam(). When I do this, gam()

Dear Trevor, I've noticed a discrepancy in the degrees of freedom reported by gam() from the gam package in R vs. gam() in S-PLUS. The nonparametric df differ by 1; otherwise (except for things that depend upon the df), the output is the same: --------- snip ------------ *** From R (gam version 0.93): > mod.gam <- gam(prestige ~ lo(income, span=.6), data=Prestige) >

Hello R Users, I have a question regarding fitting a model with GAM{mgcv}. I have data from several predictor (X) variables I wish to use to develop a model to predict one Y variable. I am working with ecological data, so have data collected many times (about 20) over the course of two years. Plotting data independently for each date there appears to be relationships between Y (fish density)

Dear list, according to the documentation of acf{ts} "the partial correlation coefficient is estimated by fitting autoregressive models of successively higher orders up to lag.max. " However, R seems to return the Yule-Walker estimates of the PACF by default. You can check this using c(1:10) as the series: the YW estimates are 0.7000000 and -0.1527035 for lags 1 and 2 . If the PACF

Recently I was using GAM and couldn't help noticing the following incoherence in prediction: > data(gam.data) > data(gam.newdata) > gam.object <- gam(y ~ s(x,6) + z, data=gam.data) > predict(gam.object)[1] 1 0.8017407 > predict(gam.object,data.frame(x=gam.data$x[1],z=gam.data$z[1])) 1 0.1668452 I would expect that using two types of predict arguments

Hi all, until now I have generally used mgcv for gams, however, I decided to experiment with the package gam, and have ran into the previously outlined problem (see below), for which I have not yet found a solution in the archives. If anyone has any suggestions, please let me know. The only difference for my case is that I am running R 2.9.0 under windows Thank you! I'm running R