R help - Nov 2009 - Interpretation of plots in linear regression models (verification of Gauss-Markov hypothesis)

An embedded and charset-unspecified text was scrubbed...
Name: not available
URL:
<https://stat.ethz.ch/pipermail/r-help/attachments/20091125/3f0e364b/attachment-0001.pl>

1) that is a loess smooth curve of the plotted points, looking at the help
?plot.lm would lead to the panel.smooth function that does the actual plotting.
2) It is described on the help page for plot.lm.
3) The basics are described on the help page.  If you don't know what a
leverage or Cook's distance is, then you probably should not be looking at
that plot, and probably not doing any real regression analyses.

You should consult a good book or other reference on regression (or at least the
regression chapter in a basic stats book) to learn these things.

-- 
Gregory (Greg) L. Snow Ph.D.
Statistical Data Center
Intermountain Healthcare
greg.snow at imail.org
801.408.8111

> -----Original Message-----
> From: r-help-bounces at r-project.org [mailto:r-help-bounces at r-
> project.org] On Behalf Of jose romero
> Sent: Wednesday, November 25, 2009 3:33 PM
> To: r-help at r-project.org
> Subject: [R] Interpretation of plots in linear regression models
> (verification of Gauss-Markov hypothesis)
> 
> Hello R Gurus:
> 
> I'm doing a simple linear regression model:
> 
> modelo1 <- lm(X9 ~ 1 + X1 + I(log(X2)) + X3 + I(log(X4)) + X5 +
> I(log(X6)) + X7)
> 
> of which i later do a plot:
> 
> plot(modelo1)
> 
> This shows 4 graphics, about which I ask:
> 
> 1) In the "Residuals vs. Fitted", what does the red curve
represent?
> 2) What does the "scale-location" graphic show? How is it
different
> from the "residuals vs. fitted? (I mean, changing the scale of the Y
> axis to show standarized residuals does not look like a big difference
> to me) What does the red curve represent in that graphic?
> 3) How do i interpret the whole "residuals vs. leverage" graphic
and
> what is that "cook's distance" business about?
> 
> I'm basically intereseted in doing a residual analysis (you know, at
> least "visually" confirming the conditions of the gauss markov
> hypothesis).?? I understand that the QQplot allows me to visually
> confirm if the residuals are normally distributed, but how do i use the
> other graphics to verify homocedasticity and independance of the
> residuals from the model variables?
> 
> Thanks in advance,
> 
> jose loreto romero
> 
> 
> 
> 
> 	[[alternative HTML version deleted]]