R help - Feb 2010 - Plot different regression models on one graph

The following variables have the following significant relationships (x is the
explanatory variable): linear, cubic, exponential, logistic. The linear
relationship plots without any trouble.

Cubic is the 'best' model, but it is not plotting as a smooth curve
using the following code:

cubic.lm<- lm(y~poly(x,3))
lines(x,predict(cubic.lm),lwd=2)

How do I plot the data and the estimated curves for all of these regression
models in the same plot?

x <- c(62.5,68.5,0,52,0,52,0,52,23.5,86,0,0,0,0,0,0,0,0,0,0)

y <-
c(0.054,0.055,0.017,0.021,0.020,0.028,0.032,0.073,0.076,0.087,0.042,0.042,0.041,0.045,0.021,0.018,0.017,0.018,0.028,0.022)

Thanks in advance.

Rhonda Reidy

	[[alternative HTML version deleted]]

Rhonda Reidy wrote:
> 
> The following variables have the following significant relationships (x is
> the explanatory variable): linear, cubic, exponential, logistic. The
> linear relationship plots without any trouble. 
> 
> Cubic is the 'best' model, but it is not plotting as a smooth curve
using
> the following code:
> 
> cubic.lm<- lm(y~poly(x,3))
> lines(x,predict(cubic.lm),lwd=2)
> 
> How do I plot the data and the estimated curves for all of these
> regression models in the same plot? 
> 
> x <- c(62.5,68.5,0,52,0,52,0,52,23.5,86,0,0,0,0,0,0,0,0,0,0)
> 
> y <-
>
c(0.054,0.055,0.017,0.021,0.020,0.028,0.032,0.073,0.076,0.087,0.042,0.042,0.041,0.045,0.021,0.018,0.017,0.018,0.028,0.022)
> 
Hi Rhonda,

The problem is that the default behavior of predict() only produces output
corresponding to the original values of x used in the model.  To get a
"smooth" curve, you will want to predict at a large number of x values
that
fall within the range of the data.  This can be done using the pretty()
function, or using a composition of the seq() and range() functions if you
desire more control:

  # Create a new vector of X values at which to generate predictions:
  predX <- pretty( x, n = 100 )

  plot( y ~ x )

  # Add the regression model, but generate predictions using the
  # points in predX-- the default is to predict using points in x.

  lines( predX, predict( cubic.lm, newdata = list( x = predX )), lwd = 2 )


The above should produce the output you want.  However, I suggest Hadley
Wickham's excellent ggplot2 package-- I feel the following is more
expressive than using base graphics functions:

  require( ggplot2 )
  Plot <- qplot( x, y ) + stat_smooth( method = 'lm', formula =
'y ~
poly(x,3)' )
  print( Plot )

The amazing power of ggplot2 is that when you want to alter or extend your
plot, you simply "add" more ggplot2 commands to your plot object:

  # Add a linear regression model and change the color theme used in the
output
  Plot <- Plot + stat_smooth( method = 'lm', color = 'red' )
+ theme_bw()
  print( Plot )


Hope this helps!

-Charlie
-- 
View this message in context:
http://n4.nabble.com/Plot-different-regression-models-on-one-graph-tp1554606p1554622.html
Sent from the R help mailing list archive at Nabble.com.

On Feb 13, 2010, at 1:35 PM, Rhonda Reidy wrote:
> The following variables have the following significant relationships  
> (x is the explanatory variable): linear, cubic, exponential,  
> logistic. The linear relationship plots without any trouble.
>
> Cubic is the 'best' model, but it is not plotting as a smooth curve
> using the following code:
>
> cubic.lm<- lm(y~poly(x,3))
Try:

lines(0:80, predict(cubic.lm, data.frame(x=0:80)),lwd=2)

But I really must say the superiority of that relationship over a  
linear one is far from convincing to my eyes. Seems to be caused by  
one data point. I hope the quotes around "best" mean tha tyou have the
same qualms.

> lines(x,predict(cubic.lm),lwd=2)
>
> How do I plot the data and the estimated curves for all of these  
> regression models in the same plot?
>
> x <- c(62.5,68.5,0,52,0,52,0,52,23.5,86,0,0,0,0,0,0,0,0,0,0)
>
> y <-  
> c 
>
(0.054,0.055,0.017,0.021,0.020,0.028,0.032,0.073,0.076,0.087,0.042,0.042,0.041,0.045,0.021,0.018,0.017,0.018,0.028,0.022
> )
>
> Thanks in advance.
>
> Rhonda Reidy
>
-- 

David Winsemius, MD
Heritage Laboratories
West Hartford, CT

Dear Rhonda,

Consider curve().

KeithC.

-----Original Message-----
From: Rhonda Reidy [mailto:rreidy at gmail.com] 
Sent: Saturday, February 13, 2010 11:36 AM
To: r-help at r-project.org
Subject: [R] Plot different regression models on one graph

The following variables have the following significant relationships (x is
the explanatory variable): linear, cubic, exponential, logistic. The linear
relationship plots without any trouble. 

Cubic is the 'best' model, but it is not plotting as a smooth curve
using
the following code:

cubic.lm<- lm(y~poly(x,3))
lines(x,predict(cubic.lm),lwd=2)

How do I plot the data and the estimated curves for all of these regression
models in the same plot? 

x <- c(62.5,68.5,0,52,0,52,0,52,23.5,86,0,0,0,0,0,0,0,0,0,0)

y <-
c(0.054,0.055,0.017,0.021,0.020,0.028,0.032,0.073,0.076,0.087,0.042,0.042,0.
041,0.045,0.021,0.018,0.017,0.018,0.028,0.022)

Thanks in advance.

Rhonda Reidy

	[[alternative HTML version deleted]]