for loop

Hi Dennis, 



Thank you for your extensive explanations. Yes, I guess I did not 

I would like to do.

Basically I would like to conduct a linear regression for each of 15 

Your answers gave me new

Perspective on how R works. 



Thanks again for the help, 

m



From: Dennis Murphy [mailto:djmuser at gmail.com] 
Sent: Sunday, October 31, 2010 4:40 AM
To: Matev?? Pavli??
Cc: r-help at r-project.org
Subject: Re: [R] for loop



Hi:

If your objective is to make 15 plots, one for each level of razred, 

don't need to make 15 individual data frames first. The lattice and 

packages allow conditioning plots. You haven't mentioned what types of 

you're interested in getting, but if it's something simple like a 

of y vs. x for each level of razred, it's not that hard to do. Let's 

d <- data.frame(razred = rep(LETTERS[1:15], each = 10),
                   x = rep(1:10, 15),
                   y = rep(2 + 0.5 * 1:10, 15) + rnorm(150))

d has 15 levels of razred with 10 observations at each level. razred is 

factor, the other variables are either integer or numeric.

Produce scatterplots of y vs. x for each level of razred, using both the 

lattice and ggplot2 packages:

library(lattice)
# each plot adds a new feature - run one plot at a time.
xyplot(y ~ x | razred, data = d, type = c('p', 'r'))
xyplot(y ~ x | razred, data = d, type = c('p', 'r'), layout = c(3, 5))
xyplot(y ~ x | razred, data = d, type = c('p', 'r'), layout = c(3, 5), 

library(ggplot2)
ggplot(d, aes(x, y)) + geom_point() + geom_smooth(method = 'lm') +
    facet_wrap( ~ razred, ncol = 3)
ggplot(d, aes(x, y)) + geom_point() + geom_smooth(method = 'lm', se = 

   facet_wrap( ~ razred, ncol = 3)

If instead you want something like a scatterplot matrix for each data 

defined by level of razred, then maybe something like this (?):

# add a new variable to the data frame
# splom is the scatterplot matrix function in lattice
d$z1 <- rnorm(150)
splom(~ d[, -1] | razred, data = d, layout = c(2, 2, 4))

Just guessing here since you didn't make your objective explicit. 

It's entirely possible that you can conduct a significant part of your 

analysis without having to split the data into subsets. Several summary 
functions, for example, can compute a number of summary functions by 

with a one-line call. Here are a couple of examples, one using 

from the base package and another using function ddply() from the plyr 

aggregate(y ~ razred, data = d, FUN = mean)
   razred        y
1       A 4.816841
2       B 4.520804
3       C 5.196329
4       D 4.615575
5       E 3.982240
6       F 4.466559
7       G 4.938669
8       H 4.539541
9       I 4.354991
10      J 4.573654
11      K 4.450624
12      L 5.138087
13      M 4.931111
14      N 4.879493
15      O 5.087452

library(plyr)
ddply(d, 'razred', summarise, mx = mean(x), my = mean(y), mz1 = 

   razred  mx       my         mz1
1       A 5.5 4.816841 -0.01745305
2       B 5.5 4.520804  0.24724069
3       C 5.5 5.196329  0.18717750
4       D 5.5 4.615575  0.18885590
5       E 5.5 3.982240 -0.91284339
6       F 5.5 4.466559  0.36479266
7       G 5.5 4.938669 -0.36359562
8       H 5.5 4.539541  0.06061162
9       I 5.5 4.354991  0.05138409
10      J 5.5 4.573654  0.31160018
11      K 5.5 4.450624  0.17458712
12      L 5.5 5.138087 -0.26482357
13      M 5.5 4.931111 -0.39194953
14      N 5.5 4.879493  0.33154075
15      O 5.5 5.087452  0.32816931

There are a number of functions and packages that will do this sort of 

quite well - I'll mention doBy, data.table, Hmisc and sqldf as excellent 

options, noting that there are other packages and functions in the apply 

family that can perform groupwise processing seamlessly. The point of 
mentioning this is so that you don't automatically think you have to 

data in myriad ways before you can process a function. The good folks 

designed this language, and the many people who have contributed code to 

project, are pretty smart, and have devised fairly simple ways to 

data, even if it's large. 

Of course, it's always possible that splitting is necessary; if you're 

to be a little more forthcoming about your analysis goals, you might get 

better targeted response..

HTH,
Dennis



On Sat, Oct 30, 2010 at 12:00 PM, Matev?? Pavli?? 

Just one more thing...
I get a list with 15 data.frames :

List of 15
 $ 1:'data.frame':      7 obs. of  9 variables:
 ..$ vrtina         : Factor w/ 6 levels "T1A-1","T1A-2",..: 1 1 2 2 5 5 

 ..$ globina.meritve: num [1:7] 7.6 8.5 10.4 17.4 12.5 15.5 16.5
 ..$ E0             : num [1:7] 4109 2533 491 810 2374 ...
 ..$ Eur1           : num [1:7] 6194 4713 605 1473 NA ...
 ..$ Eur2           : num [1:7] 3665 7216 266 4794 7387 ...
 ..$ Eur3           : num [1:7] 3221 3545 920 3347 6768 ...
 ..$ H              : num [1:7] 8 5.9 5.9 6.9 9.3 10.9 10
 ..$ Mpl            : num [1:7] 61.9 136.7 19.9 96.4 178.5 ...
 ..$ class          : int [1:7] 1 1 1 1 1 1 1
.
.
.

But how would I acces them (i.e. to draw a plot for each data.frame for 

data.frame in a list)?

Thanks,m


-----Original Message-----
From: David Winsemius [mailto:dwinsemius at comcast.net]
Sent: Saturday, October 30, 2010 8:24 PM
To: Matev?? Pavli??
Cc: r-help at r-project.org
Subject: Re: [R] for loop


On Oct 30, 2010, at 2:07 PM, Matev?? Pavli?? wrote:

Hi,

I know this is probalby a very trivial thing to do for most of the R
users, but since I just strated using it I have some problems....

I have a data.frame with a field called "razred". This field has
values from 1 up to 15.

Is it possible to create a for loop that would create a new data frame
for each of the "razred" values.

The R-way would be to use the split function and leave the result in a 

which the same operation could be also repeatedly performed using 

?split

And take a look at the fourth example applying split to the builtin 

The plyr package also provides functions on dataframes.

--

David Winsemius, MD
West Hartford, CT