Newbie question: plotting regression models

On Wed, 3 Nov 2004 10:12:53 -0500, adelmaas at MUSC.EDU wrote :

Greetings.

Is there any way to get R to take a regression model object and draw a 
plot of the regression function?  How about overlaying that plot over a 
scatterplot of the actual data?  Thanks in advance for any help anyone 
can provide.

Lots of functions in R can adapt themselves to complex objects in a
sensible way.  (These are called generic functions.)  The usual way to
draw a straight line would be to use abline(), and it can handle
linear model objects:

# fake some data
x <- 1:10
y <- rnorm(10)

# fit it and plot it.

fit <- lm(y ~ x)
plot(x, y)
abline(fit)

If you've fit a more complicated model (e.g. a quadratic), you need a
different method (because abline only works on straight lines).  Then
use

fit <- lm(y ~ x + I(x^2))
plot(x, y)
lines(predict(fit))

(This would work for the original one, too.)

Duncan Murdoch

Greetings.

Is there any way to get R to take a regression model object and draw a 
plot of the regression function?  How about overlaying that plot over a 
scatterplot of the actual data?  Thanks in advance for any help anyone 
can provide.

Aaron

Hi Aaron,

What type of "regression function" are your referring to? Linear model? 
Non-linear model? The term "regression" is to ambiguous to really answer 
your question. However, typically you should be able to do something like:

fit <- lm(y ~ x)
yhat <- predict(fit)
plot(x, y, ylim = range(c(y, yhat)))
lines(x, yhat)

If you are not using lm please provide more information than you already 
have. You should read "Introduction to R" or any of the recommended 
texts listed on the R website. Also read the posting guide.

--sundar

Welcome to R.

You can start by looking at the predict function for the regression model you are using (e.g., ?predict.lm if you are using a linear model).

Then do something like so:

data(cars)
cars.lm <- lm(dist ~ speed, cars)
dist.pred <- predict(cars.lm)
plot(cars$speed, cars$dist, ylim = c(-2, 120))
lines(cars$speed, dist.pred, col = "red", lwd = 2)


HTH, Andy

Regarding plotting a regression fit - for a simple linear regression 
model the abline function adds the fitted line to a plot.

plot(optden ~ carb, Formaldehyde)
abline(fm1 <- lm(optden ~ carb, Formaldehyde))