Hi Simon and all,
I've corrected some mistakes in setting up the prediction data frame
(sorry, very stressed and sleep deprived due to closing in deadlines),
but am having problems getting the perspective plot using persp().
I've calculated relative risk (RR) but am having trouble getting the
per
Hi Simon,
Thanks for the pointers! But I'm not sure what to do with the 'time'
variable. (I don't want to make predictions for specific points in
time). I coded as follows (full reproducible example at bottom of
email), but get a warning and error:
N <- 1000 # number of points for smooth to
On 21/07/2022 15:19, jade.shodan--- via R-help wrote:
Hello everyone (incl. Simon Wood?),
I'm not sure that my original question (see below, including
reproducible example) was as clear as it could have been. To clarify,
what I would to like to get is:
1) a perspective plot of temperature x l
Hello everyone (incl. Simon Wood?),
I'm not sure that my original question (see below, including
reproducible example) was as clear as it could have been. To clarify,
what I would to like to get is:
1) a perspective plot of temperature x lag x relative risk. I know
how to use plot.gam and vis.ga
Dear list members,
Does anyone know how to obtain a relative risk/ risk ratio from a GAM
with a distributed lag model implemented in mgcv? I have a GAM
predicting daily deaths from time series data consisting of daily
temperature, humidity and rainfall. The GAM includes a distributed lag
model bec
I have a large dataset of 118225 observations from 16 columns and as such I’ve
been using bam, rather than gam, for my analyses.
The response variable I’m using is count data but it’s overdispersed, and as
such, I thought I’d use a negative binomial model. I have 5 explanatory
variables, which
Thanks John. It's a bug in weights handling. mgcv will give wrong scale
parameter estimates for weighted models where the scale parameter is
unknown, (except Gaussian, fortunately). quasibinomial with trials > 1
is one such case, because the weights are used to store the number of
trials. Other
For both glm() and mgcv::gam() quasibinomial error models, the summary
object has a dispersion value that has the same role as sigma^2 in the
summary object for lm() model fits.
Where some fitted probabilities are small, the `gam()` default scale parameter
estimates, returned as `scale` (and `sig
@r-project.org
You can reach the person managing the list at
r-help-ow...@r-project.org
When replying, please edit your Subject line so it is more specific
than "Re: Contents of R-help digest..."
Date: Fri, 15 Mar 2019 12:31:31 +
From: Simon Wood
To: r-help@r-project.org
Sub
Can you supply the results of sessionInfo() please, and the full bam
call that causes this.
best,
Simon (mgcv maintainer)
On 15/03/2019 09:09, Frank van Berkum wrote:
> Dear Community,
>
> In our current research we are trying to fit Generalized Additive Models to a
> large dataset. We are usi
Dear Community,
In our current research we are trying to fit Generalized Additive Models to a
large dataset. We are using the package mgcv in R.
Our dataset contains about 22 million records with less than 20 risk factors
for each observation, so in our case n>>p. The dataset covers the period
Dear List,
I need to fit a GAM to a large dataset (`mgcv::bam` does this), but
ensuring that some covariates have a monotonic relation with the response.
`mgcv::mono.con` with `mgcv::pcls` seem to do this, but only for
`mgcv::gam` (not mgcv::bam)?
I'd really appreciate any pointers!
Regards,
Ax
David, Richard,
Many thanks for your responses.
Le mardi 8 janvier 2019 à 04:25:19 UTC+1, Richard M. Heiberger
a écrit :
## Here is an example using the 3-way interaction plot from the HH package
install.packages("HH") ## if necessary
## The HH package supports the book
## Statistic
## Here is an example using the 3-way interaction plot from the HH package
install.packages("HH") ## if necessary
## The HH package supports the book
## Statistical Analysis and Data Display
##Richard M. Heiberger and Burt Holland
## http://www.springer.com/us/book/9781493921218
library(HH)
On 1/7/19 3:35 PM, varin sacha via R-help wrote:
Dear R-experts,
I have fitted a model with 2-way and 3-way interactions.
I would like, for the 3-way interaction (year,age,by=education), to obtain
3D-plots. How could I do that ?
Forget ggplot2. It has ignored this sort of visualization effo
Dear R-experts,
I have fitted a model with 2-way and 3-way interactions.
I would like, for the 3-way interaction (year,age,by=education), to obtain
3D-plots. How could I do that ?
Many thanks for your response.
Here is the reproducible example:
#
install.packages("ISLR")
library
Perfect! This might be a good example to add to the documentation of mgcv
somewhere
Thanks.
Mark
On Thu, 8 Nov 2018 at 22:08, Simon Wood wrote:
> This first derivative penalty spline will do it, but the price paid is
> that the curves are often quite wiggly.
>
>
> library(mgcv); set.seed(
This first derivative penalty spline will do it, but the price paid is
that the curves are often quite wiggly.
library(mgcv); set.seed(5)
x <- runif(100); y <- x^4 + rnorm(100)*.1
b <- gam(y~s(x,m=1))
pd <- data.frame(x=seq(-.5,1.5,length=200))
ff <- predict(b,pd,se=TRUE)
plot(x,y,xlim=c(-
Dear R-help,
I have a problem where I am using the mgcv package to in a situation where
I am fitting a gam model with a 1-D spline smoother model over a domain
[a,b] but then need to make predictions and extrapolate beyond b. Is there
anyway where I force the first derivative of the spline to be z
I am having difficulty fitting a mgcv::gamm model that includes both a random
smooth term (i.e. 'fs' smooth) and autoregressive errors. Standard smooth
terms with a factor interaction using the 'by=' option work fine. Both on my
actual data and a toy example (below) I am getting the same error
I am trying to test out several mgcv::gam models in a scalar-on-function
regression analysis.
The following is the 'hierarchy' of models I would like to test:
(1) Y_i = a + integral[ X_i(t)*Beta(t) dt ]
(2) Y_i = a + integral[ F{X_i(t)}*Beta(t) dt ]
(3) Y_i = a + integral[ F{X_i(t),t} dt ]
I am trying to test out several mgcv::gam models in a scalar-on-function
regression analysis.
The following is the 'hierarchy' of models I would like to test:
(1) Y_i = a + integral[ X_i(t)*Beta(t) dt ]
(2) Y_i = a + integral[ F{X_i(t)}*Beta(t) dt ]
(3) Y_i = a + integral[ F{X_i(t),t} dt ]
Would be grateful for advice on gam/bam model selection incorporating random
effects and autoregressive terms.
I have a multivariate time series recorded on ~500 subjects at ~100 time
points. One of the variables (A) is the dependent and four others (B to E) are
predictors. My basic formula i
Dear Professor Wood,
Thank you for taking the time to fix the problem.
Best,
Fotis
On Thu, Sep 22, 2016 at 4:25 PM, Simon Wood wrote:
> Hi Fotis,
>
> Thanks for the report, and sending me the data and code (off list). The
> problem is triggered by 'ctrial' being a (one column) matrix. An imm
Hi Fotis,
Thanks for the report, and sending me the data and code (off list). The
problem is triggered by 'ctrial' being a (one column) matrix. An
immediate fix is
data_a$ctrial <- as.numeric(data_a$ctrial)
- mgcv 1.8-16 will catch the problem automatically internally.
best,
Simon
On 20/09
Any chance you could send me the data and exact code that produces this
(I'll only use the data for investigating this issue of course - often
data with the predictor replaced by noise will produce the same error,
if sending the raw data is a problem)?
best, Simon (mgcv maintainer)
On 20/09/16
Hi all
I am using the bam function of the mgcv package to model behavioral data of
a learning experiment. To model individual variation in learning rate, I am
testing models with (a) by-participant random intercepts of trial, (b)
by-participant random slopes and random intercepts of trial, and (c)
Dear Prof. Wood
Thank you, again, for your immediate response.
Best,
Fotis
On Mon, May 23, 2016 at 4:32 PM, Simon Wood wrote:
> Q1: It looks like the model is not fully identifiably given the data and
> as a result igcCAT.ideo has been set to zero - there is no sensible test to
> conduct with
Q1: It looks like the model is not fully identifiably given the data and
as a result igcCAT.ideo has been set to zero - there is no sensible test
to conduct with such a term, hence the NAs in the test stat an p-value
fields.
Q2: A separate (centred) smooth is estimated for each level of igc. I
Hallo all
I am using a gam model for my data.
m2.4<-bam(acc~ 1 + igc + s(ctrial, by=igc) + shape + s(ctrial, by=shape) +
s(ctrial, sbj, bs = "fs", m = 1) , data=data, family=binomial)
igc codes condition and there are four levels (CAT.pseudo,
CAT.ideo,PA.pseudo, PA.ideo), and shape is a factor (
Hi Simon,
Thanks for getting back to me. Would the opposite be true as well then? I
have other models where I don't get a warning message at the end of the
model fitting but the $converged slow indicates a 'FALSE'.
Thank you ~Trevor
On Wed, Jun 18, 2014 at 2:07 AM, Simon Wood wrote:
> I thin
Hello Simon,
thanks for your response!
I incorporated the intercept, i.e. coef(my_gam)[1] into the spline. That
is, my B-spline coefficients are the intercept for the first one, and
for all others I added the intercept to them (see the code below if I'm
not clear)
Now I *almost* reproduce t
I think it didn't converge. The warning message is for the whole fitting
iteration, whereas mod07.3.BASE.bam5.1$converged indicates whether
smoothing parameter selection converged at the final step of the iteration.
best,
Simon
On 17/06/14 19:10, Trevor Davies wrote:
I'm running some spatial
mgcv:gam automatically imposes sum-to-zero constraints on the smooths in
a model (even if there is only one smooth). This is to avoid lack of
identifiability with the intercept. The constraint removes one
coefficient and shifts the curve...
best,
Simon
On 17/06/14 15:40, Alexander Engelhardt
I'm running some spatial GAMs and am using the bam call (negbin family) and
am getting conflicting information on whether the model is converging or
not. When the model completes its run, I get a warning message that the
model did not converge. When I look at the object itself, I'm told that it
d
Hello R-helpers,
I am working through Simon Wood's GAM book and want to specify my own
knot locations (on even tens, i.e. 10, 20, 30, etc.). Then, I want to
compute a GAM on that area, and given the coefficients, reconstruct the
same P-spline that is drawn in plot(my_gam).
I'm failing.
Here
Dear R-helpers,
I am working on a project assessing the prevalence and variance (random
effects) of linear and nonlinear trends in a data set that has short
time series (each time series identified as PID 1 through 5). I am using
mgcv (gam) with the bs=”re” option (more on why not gamm or gamm
Hi Michael,
You seem to have a quite recent mgcv if you are using qq.gam, so a help
file for version 1.3 is probably not going to be much help (gam.fit2 no
longer exists, for example).
By default qq.gam plots deviance residuals (see ?qq.gam). So the default
is standardization. When possible
Hello all,
I looking for confirmation of what I'm seeing. The qq plot in gam.check
and qq.gam is not standardizing the residuals.
The current help doesn't suggest they're standardized.
Somehow I found, online, a help for gam.check from version [Package
mgcv version 1.3-23 Index]:
"If the fit
Dear Dr. Wood and other mgcv experts
In ?gam.models, it says that the numeric "by" variable is genrally not
subjected to an identifiability constraint, and I used the example in
?gam.models, finding some differences (code below).
I think the the problem might become serious when several varying
Dear Prof Ripley
yes, but if the estimate is biased it's good to know what the bias is.
The problem illustrated in the simulations has nothing to do with ML,
though, as the default fitting method in mgcv when scale is unknown is
"GCV" and that is what was used, by default, here.
The point about
hi Simon
yes, I also got the right shape of the mean-variance relation but the
wrong estimate of the parameter.
thanks very much
Greg
> Hi Greg,
>
> Yes, this sounds right - with quasipoisson gam uses `extended
> quasi-likelihood' (see McCullagh and Nelder's GLM book) to allow
> estimation of t
On 05/02/2014 12:56, Greg Dropkin wrote:
thanks Simon
also, it appears at least with ML that the default scale estimate with
quasipoisson (i.e. using dev) is the scale which minimises the ML value of
the fitted model. So it is the "best" model but doesn't actually give the
correct mean-variance
thanks Simon
also, it appears at least with ML that the default scale estimate with
quasipoisson (i.e. using dev) is the scale which minimises the ML value of
the fitted model. So it is the "best" model but doesn't actually give the
correct mean-variance relation. Is that right?
thanks again
Gre
Hi Greg,
Yes, this sounds right - with quasipoisson gam uses `extended
quasi-likelihood' (see McCullagh and Nelder's GLM book) to allow
estimation of the scale parameter along with the smoothing parameters
via (RE)ML, and it could well be that this gives a biased scale estimate
with low count
Greg,
The deviance being chi^2 distributed on the residual degrees of freedom
works in some cases (mostly where the response itself can be reasonably
approximated as Gaussian), but rather poorly in others (noteably low
count data). This is what you are seeing in your simulations - in the
firs
mgcv: distribution of dev
hi
I can't tell if this is a simple error.
I'm puzzled by the distribution of dev when fitting a gam to Poisson
generated data.
I expected dev to be approximately chi-squared on residual d.f., i.e.
about 1000 in each case below.
In particular, the low values in the 3rd a
Dear R-users,
Happy new year to all!
I have been using the mgcv package, and I have run some models using the
option mrf, for saptial data. But I have found quite hard to interpret the
results. I could not find a lot of documentation on that, examples and so
on, so I was wondering if anyone can h
Dear R-Helpers,
I posted two days ago on testing significance of random effects in mgcv,
but realize I did not make my overall purpose clear. I have a series of
N short time series, where N might range from 3-10 and short means a
median of 20 time points. The sample data below (PCP) has N = 4
Dear useR,
I don't understand the results of the predict.bam function of mgcv package
when constucting a varying-coefficient model with bam instead of gam:
library("mgcv")
dat <- gamSim(4)
b <- gam(y ~ fac+s(x2,by=fac)+s(x0), data=dat)
predict(b, dat[1,], type = "terms")
with gam everything is
On Tue, 2013-07-23 at 11:16 +0200, Christoph Scherber wrote:
> Dear all,
>
> This is just a quick question regarding degrees of freedom in GAM
> models fit by MGCV (using select=T):
>
> Can I roughly interpret them as:
>
> 1 df: linear effect of x on y
> 2 df: approximately quadratic of x on y
>
I can't see anything immediately wrong except:
1. presumably there are repeated values in 'road_quiet' aren't there? If
so then your inequality constraint matrix
will contain constraints that are *exact* copies of each other. I'm not
sure, and don't immediately have time to try it
out, but it c
Dear R helplisters,
I am trying to implement a mononicity constraint on a smooth term in my
generalized additive model with the mgcv package (v. 1.7-18). I adapted the
code from an example given in the help file for pcls(). The example runs just
as one would expect, but when I adapt it and use
Dear all,
This is just a quick question regarding degrees of freedom in GAM models fit by
MGCV (using select=T):
Can I roughly interpret them as:
1 df: linear effect of x on y
2 df: approximately quadratic of x on y
3 df: approximately cubic effect of x on y?
1 df for a spatial term s(x,y): bil
Dear R help list,
This is a long post so apologies in advance. I am estimating a model with the
mgcv package, which has several covariates both linear and smooth terms. For 1
or 2 of these smooth terms, I "know" that the truth is monotonic and downward
sloping. I am aware that a new package "s
Meanwhile, I found the solution myself:
using plot.gam() with shift=intercept and trans=exp (for a poisson model) does
the job. I can then
add the original data using points()
Thanks again for your help, which is greatly appreciated!
Best wishes
Christoph
Am 16/07/2013 11:04, schrieb Simon Woo
Thanks, the sequence of x0 values was clearly too short.
However, is there a way to overlay the (marginal) curve from plot.gam() over a
plot of (x,y) values?
Best wishes
Christoph
Am 16/07/2013 11:04, schrieb Simon Wood:
> Probably you didn't want to set x0=0:1? Here is some code, to do what
Probably you didn't want to set x0=0:1? Here is some code, to do what you want.
(The CI shape is not identical to the plot(b) version as the uncertainty
includes
the uncertainty in the other smooths and the intercept now.)
library(mgcv)
set.seed(2)
dat <- gamSim(1,n=400,dist="normal",scale=2)
b
Dear R users,
I´ve stumbled over a problem that can be easily seen from the R code below:
- When I use plot.gam() on a fitted model object, I get a nice and well-looking
smooth curve for all
terms in the model.
- However, when I use predict(model) for a given predictor, with values of all
othe
I think it's going to be a problem to have different sized groups in
your second model. ?corSymm says that a general correlation matrix is
being estimated (i.e. the correlation between each pair of observations
is being estimated - for this to be meaningful across groups you need
the jth price
Dear Help list,
I am relatively new to the mgcv package, which I am using to model prices of
housing transactions as a function of the characteristics of a home and a
neighborhood. I have several smooth terms to capture price evolution over time,
but also to non-parametrically fit the functiona
hi
I'm trying to understand (a little) the code behind summary.gam, and have
the Biometrika article referred to in the Help. But am stuck early on.
In the code, starting at line 167:
if (est.disp)
rdf <- residual.df
else rdf <- -1
res <- testStat(p, Xt, V, df[i], type = p.type,
res.df = rdf)
Hi Greg
On 21/06/13 08:02, Greg Dropkin wrote:
> hi
>
> I'm trying to understand (a little) the code behind summary.gam, and have
> the Biometrika article referred to in the Help. But am stuck early on.
>
> In the code, starting at line 167:
>
> if (est.disp)
>rdf <- residual.df
> else rdf <-
Juliet,
for you the diagnostic plots:
just to recall:
the first model was this:
fit<-gam(target
~s(mgs)+s(gsd)+s(mud)+s(ssCmax),family=quasi(link=log),data=wspe1,method="REML",select=F)
> summary(fit)
Parametric coefficients:
Estimate Std. Error t value Pr(>|t|)
Hi Jan and Simon,
If possible, could you attach the diagnostic plots. I would be curious to
see them.
Thanks,
Juliet
On Fri, Apr 19, 2013 at 4:39 AM, jholstei wrote:
> Simon,
>
> that was very instructivevery special thanks to you.
> I already noticed that the model was bad, but it was not
Simon,
that was very instructivevery special thanks to you.
I already noticed that the model was bad, but it was not clear to me that
transformation of predictors to, say a more centered distribution is helpful
here.
And thanks for pointing out Tweedie, I noticed that the error structure is fa
Jan,
Thanks for the data (off list). The p-value computations are based on
the approximation that things are approximately normal on the linear
predictor scale, but actually they are no where close to normal in this
case, which is why the p-values look inconsistent. The reason that the
approx
Jan,
Thanks for this. Is there any chance that you could send me the data off
list and I'll try to figure out what is happening? (Under the
understanding that I'll only use the data for investigating this issue,
of course).
best,
Simon
on 18/04/13 11:11, Jan Holstein wrote:
Simon,
thanks
Simon,
thanks for the reply, I guess I'm pretty much up to date using
mgcv 1.7-22.
Upgrading to R 3.0.0 also didn't do any change.
Unfortunately using method="REML" does not make any difference:
### first with "select=FALSE"
> fit<-gam(target
> ~s(mgs)+s(gsd)+s(mud)+s(ssCmax),family=quasi(
Jan,
What mgcv version are you using, please? (Older versions have a poor
p-value approximation when select=TRUE, but of course it's possible that
you've managed to break the newer approximation as well)
The 'select=TRUE' option adds a penalty to each smooth, to allow it to
be penalized out
I have 11 possible predictor variables and use them to model quite a few
target variables.
In search for a consistent manner and possibly non-manual manner to identify
the significant predictor vars out of the eleven I thought the option
"select=T" might do.
Example: (here only 4 pedictors)
firs
Solved my own problem. If interested:
http://stackoverflow.com/questions/15563589/odd-error-error-in-predictmatobjectsmoothk-data-by-variable-must
On 03/21/2013 02:14 PM, Andrew Crane-Droesch wrote:
Dear List,
I'm getting an error in mgcv, and I can't figure out where it comes
from. The se
Dear List,
I'm getting an error in mgcv, and I can't figure out where it comes
from. The setup is the following: I've got a fitted GAM object called
"MI", and a vector of "prediction data" (with default values for
predictors). I feed this into predict.gam(object, newdata = whatever)
via th
mgcv: Constructing probabilities for binomial GAM with crossed random
intercepts and factor by variable
Hello,
(I'm sorry if this has been discussed elsewhere; I may not have been
looking in the right places.)
I ran a binomial GAM in which "Correct" is modelled in terms of the
participant's
plot(fit,all.terms=TRUE,select=3,xlabs="???")
On 20/12/12 02:06, Andrew Crane-Droesch wrote:
Dear List,
plot.gam appears to be having trouble communicating its xlab and ylab
information to termplot. A simple example:
library(mgcv)
x = 1:1000
y = runif(1000)*x^.5
z = rnorm(1000)*y
other = sin(
Dear List,
plot.gam appears to be having trouble communicating its xlab and ylab
information to termplot. A simple example:
library(mgcv)
x = 1:1000
y = runif(1000)*x^.5
z = rnorm(1000)*y
other = sin(z)
fit = gam(y~s(x)+s(z)+other)
plot(fit,all.terms=TRUE)
plot(fit,select=3,xlab="???")
de
Andrew,
I think you mean
dumb.example2$coefficients = averaged.models
~~~
currently your code adds a new vector 'coeff' to the gam object, rather
than modifying 'coefficients'. (A good example of why forgiving
languages like R are dangerous, and you should really wr
Hi Simon,
Thanks for your help. I've got another question if you don't mind -- is
it possible to "swap out" a set of coefficients of a gamObject in order
to change the results when that gamObject is plotted? The (silly)
example below illustrates that this is possible with the Vp matrix. But
Hi Andrew,
mgcv matches the knots to the smooth arguments by name. If an element of
'knots' has
no name it will be ignored. The following will do what you want...
dumb.example = gam(y~s(x,k=3),knots=list(x=dumb.knots))
best,
Simon
On 29/11/12 23:44, Andrew Crane-Droesch wrote:
Dear List,
I'
Dear List,
I'm using GAMs in a multiple imputation project, and I want to be able
to combine the parameter estimates and covariance matrices from each
completed dataset's fitted model in the end. In order to do this, I
need the knots to be uniform for each model with partially-imputed
data.
Hi Simon,
I'm trying to fit a negative binomial gam with no covariates, that therefore
looks at the detection/non-detection data and nothing else. I thought having
1~1 as formula would allow the model to just estimate occurrence without
looking at relationship between variables.
On 25 Aug 20
'gam' doesn't know what to do with the model formula '1 ~ 1' (i.e. "one
tilde one"). What is it supposed to mean? 'glm' also does nothing
meaningful in this case...
## code to load you data file into 'dat' omitted
> model<-glm(1~1, data=dat)
> model
Call: glm(formula = 1 ~ 1, data = dat)
Co
I've attached the data to this message.
http://r.789695.n4.nabble.com/file/n4641291/2005pip.csv 2005pip.csv
--
View this message in context:
http://r.789695.n4.nabble.com/mgcv-package-problems-with-NAs-in-gam-tp4641253p4641291.html
Sent from the R help mailing list archive at Nabble.com.
Grace:
Confession: I loved that error message! -- and it seems pretty clear to me.
What does "to no avail" mean -- in particular, what happened when you
changed your NA's to 0? Presumably you did not get the same error
message, again, but something else, right? What else?
Modulo the above vague
Hi there,
I'm using presence-absence data in a gam (i.e. 0 or 1 as values)
I am trying to run a gam with 'dummy covariates' i.e. 1~1
unfortunately my model:
*
model<-gam(1~1, data=bats, family=negbin)*
keeps putting out:
*
Error in gam(1 ~ 1, data = bats, family = negbin) :
Not enough (non-NA
john benson hotmail.com> writes:
> I've been using gamm4 to build GAMMs for exploring environmental
> influences on genetic ancestry. Things have gone well and I have 2
> very straightforward questions: 1. I've used method=REML. Am I
> correct that this is an alternative method for estimating
Hi,
I've been using gamm4 to build GAMMs for exploring environmental influences on
genetic ancestry. Things have gone well and I have 2 very straightforward
questions:
1. I've used method=REML. Am I correct that this is an alternative method for
estimating the smooth functions in GAMMs rather
I somehow solved the problem - kind of. The data set on which I ran the GAM
model contains many more variables than are needed in the model, so I
created a new data set in R and reran the GAM model on the slimmed down data
set. Same problem: The GAM can be computed, but the tensor product cannot be
Hi Simon,
Thanks for your reply.
m <- bam(Correct ~ cEnglishTotal + te(WSTResid, RavenResid) + s(Stimulus,
bs="re") + s(Subject, bs="re"), data = dat, family = "binomial")
# cEnglishTotal, WSTResid and RavenResid are continuous variables; Correct,
Stimulus and Subject are factors.
> vis.gam(m, v
Jan,
Could you send the exact gam call and exact vis.gam call that did this
please? Also, if 'm' denotes your fitted model, what result does
'fitted(m)' give? and what is the output from print(m)?
best,
Simon
On 07/30/2012 09:19 PM, janvanhove wrote:
Hi everyone,
I ran a binomial GAM consi
Hi everyone,
I ran a binomial GAM consisting of a tensor product of two continuous
variables, a continuous parametric term and crossed random intercepts on a
data set with 13,042 rows. When trying to plot the tensor product with
vis.gam(), I get the following error message:
Error in persp.default
Hi Jan ,
coef(mod) will extract the coefficients, including for a. They are
labelled and for 'a' are in the same order as levels(a).
best,
Simon
On 23/07/12 10:08, janvanhove wrote:
Hi everyone,
I can't figure out how to extract by-factor random effect adjustments from a
gam model (mgcv p
Hi everyone,
I can't figure out how to extract by-factor random effect adjustments from a
gam model (mgcv package).
Example (from ?gam.vcomp):
library(mgcv)
set.seed(3)
dat <- gamSim(1,n=400,dist="normal",scale=2)
a <- factor(sample(1:10,400,replace=TRUE))
b <- factor(sample(1:7,400,replace=TRUE)
Martin,
I've just submitted mgcv_1.7-19 to CRAN, which includes a major upgrade
of the p-value computation for random effect terms (and any other smooth
term which can be penalized to zero as part of estimation). The new
p-values are still conditional on the smoothing parameter/variance
compo
Martin,
I had a nagging feeling that there must be more to this than my previous
reply suggested, and have been digging a bit further. Basically I would
expect these p-values to not be great if you had nested random effects
(such as main effects + their interaction), but your case looked rathe
Hi,Dear Professor Wood,
I am studying the generalized additive model(GAM)
and I have read your related papers,especially the book named "Generalized
Additive Models:An Introduction with R".I am benefit a lots.However,I also
have some questions about GAM models.
First,Is there some restriction
Hi Simon,
Thanks for taking the time to reply. Please let me explain a few more details.
The problem that I am working on is essentially the same as the
Bristol Channel Sole Egg distribution example in your book and in the
"soap" paper but instead it is Herring Larvae in the English Channel -
sam
Hi Mark,
irls.reg is kind of `legacy code'. Does model fitting actually fail for
your example, or is it just that the
estimated spatial smooth looks unpleasant?
best,
Simon
On 06/21/2012 01:28 AM, r-help.20.tre...@spamgourmet.com wrote:
Hi,
In the help files in the mgcv package for the ga
Hi,
In the help files in the mgcv package for the gam.control() function,
there is an option irls.reg. The help files describe this option as:
For most models this should be 0. The iteratively re-weighted least squares
method by which GAMs are fitted can fail to converge in some circumstances.
F
Hi Martijn,
Irrespective of the p-value, 'bam' and 'lmer' agree that the variance
component for 'Placename' is practically zero. In the 'bam' output see
the 'edf' for s(Placename), or for a more direct comparison call
gam.vcomp(m1).
As mentioned in ?summary.gam the p-values for "re" terms ar
1 - 100 of 181 matches
Mail list logo
