Submitted by dylan on Mon, 2009-04-27 15:47.
Submitted by dylan on Tue, 2009-02-17 04:43.
Small update to a similar thread from last week, on the comparison of slope and intercept terms fit to a multi-level model. I finally figured out (thanks R-Help mailing list!) how to efficiently use contrasts in R. The C() function can be called within a model formula, to reset the base level of an un-ordered factor. The UCLA Stats Library has an extensive description of this topic here. This approach can be used to sequentially test for differences between slope and intercept terms from a multi-level model, by re-setting the base level of a factor. See example data and figure below.
Note that the multcomp package has a much more robust approach to this type of operation. Details below.
# need these
# replicate an important experimental dataset
x <- rnorm(100)
y1 <- x[1:25] * 2 + rnorm(25, mean=1)
y2 <- x[26:50] * 2.6 + rnorm(25, mean=1.5)
y3 <- x[51:75] * 2.9 + rnorm(25, mean=5)
y4 <- x[76:100] * 3.5 + rnorm(25, mean=5.5)
d <- data.frame(x=x, y=c(y1,y2,y3,y4), f=factor(rep(letters[1:4], each=25)))
xyplot(y ~ x, groups=f, data=d,
auto.key=list(columns=4, title='Beard Type', lines=TRUE, points=FALSE, cex=0.75),
type=c('p','r'), ylab='Number of Pirates', xlab='Distance from Land')
Submitted by dylan on Thu, 2009-01-29 18:23.
When the relationship between two variable is (potentially) dependent on a third, categorical variable ANCOVA (analysis of covariance), or some variant, is commonly used. There are several approaches to testing for differences in slope/intercepts (in the case of a simple linear model) between levels of the stratifying variable. In R the following formula notation is usually used to test for interaction between levels of a factor (f) and the relationship between two continuous variables x and y: y ~ x * f. A simple graphical exploration of this type of model can be done through examination of confidence intervals computed for slope and intercept terms, for each level of our grouping factor (f). An example of a fictitious dataset is presented below. Note that this a rough approximation for testing differences in slope/intercept within a multi-level model. A more robust approach would take into account that we are trying to make several pair-wise comparisons, i.e. something akin to Tukey's HSD. Something like this can be done with the multcomp package. For any real data set you should always consult a real statistician.
# need this for xyplot()
# make some fake data:
x <- rnorm(100, mean=3, sd=6)
y <- x * runif(100, min=1, max=7) + runif(100, min=1.8, max=5)
d <- data.frame(x, y, f=rep(letters[1:10], each=10))
# check it out
xyplot(y ~ x | f, data=d, type=c('p','r'))
Submitted by dylan on Thu, 2009-01-15 04:36.
Wanted to make something akin to an interpolated surface for some spatially auto-correlated categorical data (presence/absence). I quickly generated some fake spatially auto-correlated data to work with using r.surf.fractal in GRASS. These data were converted into a binary map using an arbitrary threshold that looked about right-- splitting the data into something that looked 'spatially clumpy'.
I had used voronoi polygons in the past to display connectivity of categorical data recorded at points, even though sparsely sampled areas tend to be over emphasized. Figure 1 shows the fake spatially auto-correlated data (grey = presence /white = not present), sample points (yellow boxes), and voronoi polygons. The polygons with thicker, red boundaries represent the "voronoi interpolation" of the categorical feature.
Submitted by dylan on Tue, 2008-12-30 02:45.
Submitted by dylan on Fri, 2008-12-26 21:26.
Submitted by dylan on Tue, 2008-12-23 01:46.
Submitted by dylan on Sat, 2008-12-20 06:44.
Submitted by dylan on Mon, 2008-10-27 16:08.
Submitted by dylan on Wed, 2008-10-08 20:13.