Create a gist now

Instantly share code, notes, and snippets.

What would you like to do?
set.seed(2)
x <- 1:100
y <- 20 + 3 * x
e <- rnorm(100, 0, 60)
y <- 20 + 3 * x + e
plot(x,y)
yx.lm <- lm(y ~ x)
lines(x, predict(yx.lm), col="red")
xy.lm <- lm(x ~ y)
lines(predict(xy.lm), y, col="blue")
# so lm() depends on which variable is x and wich is y
# lm minimizes y distance (the error term is y-yhat)
#normalize means and cbind together
xyNorm <- cbind(x=x-mean(x), y=y-mean(y))
plot(xyNorm)
#covariance
xyCov <- cov(xyNorm)
eigenValues <- eigen(xyCov)$values
eigenVectors <- eigen(xyCov)$vectors
eigenValues
eigenVectors
plot(xyNorm, ylim=c(-200,200), xlim=c(-200,200))
lines(xyNorm[x], eigenVectors[2,1]/eigenVectors[1,1] * xyNorm[x])
lines(xyNorm[x], eigenVectors[2,2]/eigenVectors[1,2] * xyNorm[x])
# the largest eigenValue is the first one
# so that's our principal component.
# but the principal component is in normalized terms (mean=0)
# and we want it back in real terms like our starting data
# so let's denormalize it
plot(x,y)
lines(x, (eigenVectors[2,1]/eigenVectors[1,1] * xyNorm[x]) + mean(y))
# that looks right. line through the middle as expected
# what if we bring back our other two regressions?
lines(x, predict(yx.lm), col="red")
lines(predict(xy.lm), y, col="blue")

rubgb commented Sep 17, 2010

Hi there,
this is a great post. I have always wondered about the intuition behind PCA...
I'm curious about two of your graphs posted in the blog illustrating the difference between errors in yx and xy.
Would it be possible to have access to the R code to generate them?
Many thanks in advance,
Ruben

It's in the first 13 lines. Plot the lines() one at a time to recreate.

rubgb commented Sep 17, 2010

I meant the code for generating:
OLS1.png
OLS2.png
pca.png

Many thanks in advance,
Ruben

Ahh. I just had a revelation in what you were asking. The yellow lines in all 3 of those were drawn by hand. I did the base plots (which I think you can see from the code) and then I drew 2 yellow lines on each one for illustration.

Does that answer your question? Can you pick out of the code where the base plots are?

plot(x,y)
yx.lm <- lm(y ~ x)
lines(x, predict(yx.lm), col="red") # <- that's OLS1

plot(x,y)
xy.lm <- lm(x ~ y)
lines(predict(xy.lm), y, col="blue") # <- that's OLS2

rubgb commented Sep 17, 2010

Nice trick ha,ha. Very illustrative!
Now I'm curious. Would it be possible to do something like that with R?

And Josh Ulrich provided an answer in under 20 minutes. Hive mind FTW!

http://stackoverflow.com/questions/3737165/drop-lines-from-actual-to-modeled-points-in-r/3737183#3737183

-JD

rubgb commented Sep 18, 2010

Thanks a lot!

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment