R › R help

fitting several lme sistematically

Classic

List

Threaded

4 messages Options

Berta Ibáñez

fitting several lme sistematically

Dear R-list,

I have a data set (in the following example called "a") which have:

one "subject indicator" variable (called "id")
three dependent variables (varD, varE, var F)
three independent variables (varA, varB, varC)

I want to fit 9 lme models, one per posible combination (DA, DB, DC, EA, EB, EC, FA, FB, FC).
In stead of writting the 9 lme models, I want to do it sistematically (the example is a simplification of what I really have). Here you have the comands for the first model:

library(nlme)
set.seed(50)
a<-data.frame(array(c(rep(1:10,10), rnorm(600)), c(100,7)))
names(a)<-c("id", "varA", "varB", "varC", "varD", "varE", "varF")
lme(varD ~ varA , random= ~1|id, data=a, na.action="na.exclude")

I supossed that a simple sintaxis going through the variables of dataset "a" could cope with it:

for(i in 2:4){
for(j in 5:7){
lme(a[,j] ~ a[,i] , random= ~1|id, data=a, na.action="na.exclude")
}}

but it does not, and the use of eval, as.symbol and so on does not help.

for(i in 2:4){
for(j in 5:7){
lme(eval(as.symbol(names(a)[j])) ~ eval(as.symbol(names(a)[i])) , random= ~1|id, data=a, na.action="na.exclude")
}}

Any help??? Thanks a lot in advance!

[[alternative HTML version deleted]]

______________________________________________
[hidden email] mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.

Adams, Jean

Re: fitting several lme sistematically

I'm not sure why, but lme() doesn't seem to like the variables to be
referenced as part of a list using [ or $.
Here's an easy workaround ...

ids <- a$id
for(i in 2:4){
for(j in 5:7){
y <- a[, j]
x <- a[, i]
lme(y ~ x , random= ~1|ids, na.action="na.exclude")
}}

Jean

Berta Ibáñez <[hidden email]> wrote on 07/18/2012 08:53:51 AM:

> Dear R-list,
>
> I have a data set (in the following example called "a") which have:
>
> one "subject indicator" variable (called "id")
> three dependent variables (varD, varE, var F)
> three independent variables (varA, varB, varC)
>
> I want to fit 9 lme models, one per posible combination (DA, DB, DC,
> EA, EB, EC, FA, FB, FC).
> In stead of writting the 9 lme models, I want to do it
> sistematically (the example is a simplification of what I really
> have). Here you have the comands for the first model:
>
> library(nlme)
> set.seed(50)
> a<-data.frame(array(c(rep(1:10,10), rnorm(600)), c(100,7)))
> names(a)<-c("id", "varA", "varB", "varC", "varD", "varE", "varF")
> lme(varD ~ varA , random= ~1|id, data=a, na.action="na.exclude")
>
> I supossed that a simple sintaxis going through the variables of
> dataset "a" could cope with it:
>
> for(i in 2:4){
> for(j in 5:7){
> lme(a[,j] ~ a[,i] , random= ~1|id, data=a, na.action="na.exclude")
> }}
>
> but it does not, and the use of eval, as.symbol and so on does not help.

>
> for(i in 2:4){
> for(j in 5:7){
> lme(eval(as.symbol(names(a)[j])) ~ eval(as.symbol(names(a)[i])) ,
> random= ~1|id, data=a, na.action="na.exclude")
> }}
>
> Any help??? Thanks a lot in advance!

[[alternative HTML version deleted]]

______________________________________________
[hidden email] mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.

arun kirshna

Re: fitting several lme sistematically

Hi Jean,

Is there something missing in the function?

ids <- a$id
for(i in 2:4){
for(j in 5:7){
        y <- a[, j]
        x <- a[, i]
        model<-lme(y ~ x , random= ~1|ids, na.action="na.exclude")
        }}

summary(model)
Linear mixed-effects model fit by REML
Data: NULL
       AIC      BIC    logLik
281.1838 291.5236 -136.5919

Random effects:
Formula: ~1 | ids
        (Intercept) Residual
StdDev:   0.1109054 0.9251637

Fixed effects: y ~ x
                  Value Std.Error DF    t-value p-value
(Intercept) 0.03931479 0.09909825 89 0.3967254 0.6925
x           -0.11826047 0.09731719 89 -1.2152063 0.2275
Correlation:
(Intr)
x 0.056

Standardized Within-Group Residuals:
       Min         Q1        Med         Q3        Max
-2.0882452 -0.7718563 0.1156507 0.6119178 1.7986478

Number of Observations: 100
Number of Groups: 10

A.K.

----- Original Message -----
From: Jean V Adams <[hidden email]>
To: Berta Ibáñez <[hidden email]>
Cc: Lista de R <[hidden email]>
Sent: Wednesday, July 18, 2012 1:02 PM
Subject: Re: [R] fitting several lme sistematically

I'm not sure why, but lme() doesn't seem to like the variables to be
referenced as part of a list using [ or $.
Here's an easy workaround ...

ids <- a$id
for(i in 2:4){
for(j in 5:7){
y <- a[, j]
x <- a[, i]
lme(y ~ x , random= ~1|ids, na.action="na.exclude")
}}

Jean

Berta Ibáñez <[hidden email]> wrote on 07/18/2012 08:53:51 AM:

>
> for(i in 2:4){
> for(j in 5:7){
> lme(eval(as.symbol(names(a)[j])) ~ eval(as.symbol(names(a)[i])) ,
> random= ~1|id, data=a, na.action="na.exclude")
> }}
>
> Any help??? Thanks a lot in advance!

[[alternative HTML version deleted]]

______________________________________________
[hidden email] mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.

______________________________________________
[hidden email] mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.

Adams, Jean

Re: fitting several lme sistematically

You have only saved the last run of the for() loop (when i is 4 and j is
7) to the object called model. If you want to save all of the lme fits,
you need to set give some dimensionality to the object model. For
example, you could define model as a list, then store the lme fits as
separate elements of the list.

ids <- a$id
model <- vector("list", 9)
count <- 0
for(i in 2:4){
for(j in 5:7){
count <- count + 1
y <- a[, j]
x <- a[, i]
model[[count]] <- lme(y ~ x , random= ~1|ids,
na.action="na.exclude")
}}

Now you can summarize any of the fits, or get predicted values, or
whatever ...

summary(model[[5]])
predict(model[[7]])

Jean

arun <[hidden email]> wrote on 07/18/2012 12:34:31 PM:

> Hi Jean,
>
> Is there something missing in the function?
>
>
> ids <- a$id
> for(i in 2:4){
> for(j in 5:7){
> y <- a[, j]
> x <- a[, i]
> model<-lme(y ~ x , random= ~1|ids, na.action="na.exclude")
> }}
>
> summary(model)
> Linear mixed-effects model fit by REML
> Data: NULL
> AIC BIC logLik
> 281.1838 291.5236 -136.5919
>
> Random effects:
> Formula: ~1 | ids
> (Intercept) Residual
> StdDev: 0.1109054 0.9251637
>
> Fixed effects: y ~ x
> Value Std.Error DF t-value p-value
> (Intercept) 0.03931479 0.09909825 89 0.3967254 0.6925
> x -0.11826047 0.09731719 89 -1.2152063 0.2275
> Correlation:
> (Intr)
> x 0.056
>
> Standardized Within-Group Residuals:
> Min Q1 Med Q3 Max
> -2.0882452 -0.7718563 0.1156507 0.6119178 1.7986478
>
> Number of Observations: 100
> Number of Groups: 10
>
> A.K.
>
>
>
> ----- Original Message -----
> From: Jean V Adams <[hidden email]>
> To: Berta Ibáñez <[hidden email]>
> Cc: Lista de R <[hidden email]>
> Sent: Wednesday, July 18, 2012 1:02 PM
> Subject: Re: [R] fitting several lme sistematically
>
> I'm not sure why, but lme() doesn't seem to like the variables to be
> referenced as part of a list using [ or $.
> Here's an easy workaround ...
>
> ids <- a$id
> for(i in 2:4){
> for(j in 5:7){
> y <- a[, j]
> x <- a[, i]
> lme(y ~ x , random= ~1|ids, na.action="na.exclude")
> }}
>
> Jean
>
>
> Berta Ibáñez <[hidden email]> wrote on 07/18/2012 08:53:51 AM:
>
> > Dear R-list,
> >
> > I have a data set (in the following example called "a") which have:
> >
> > one "subject indicator" variable (called "id")
> > three dependent variables (varD, varE, var F)
> > three independent variables (varA, varB, varC)
> >
> > I want to fit 9 lme models, one per posible combination (DA, DB, DC,
> > EA, EB, EC, FA, FB, FC).
> > In stead of writting the 9 lme models, I want to do it
> > sistematically (the example is a simplification of what I really
> > have). Here you have the comands for the first model:
> >
> > library(nlme)
> > set.seed(50)
> > a<-data.frame(array(c(rep(1:10,10), rnorm(600)), c(100,7)))
> > names(a)<-c("id", "varA", "varB", "varC", "varD", "varE", "varF")
> > lme(varD ~ varA , random= ~1|id, data=a, na.action="na.exclude")
> >
> > I supossed that a simple sintaxis going through the variables of
> > dataset "a" could cope with it:
> >
> > for(i in 2:4){
> > for(j in 5:7){
> > lme(a[,j] ~ a[,i] , random= ~1|id, data=a, na.action="na.exclude")
> > }}
> >
> > but it does not, and the use of eval, as.symbol and so on does not

help.
>
> >
> > for(i in 2:4){
> > for(j in 5:7){
> > lme(eval(as.symbol(names(a)[j])) ~ eval(as.symbol(names(a)[i])) ,
> > random= ~1|id, data=a, na.action="na.exclude")
> > }}
> >
> > Any help??? Thanks a lot in advance!