Message-ID: <43F88B73.4000704@yandex.ru>
Date: 2006-02-19T15:14:59Z
From: Evgeniy Kachalin
Subject: Two factors -> nurical data dependency analyzing
In-Reply-To: <Pine.LNX.4.64.0602191356450.20648@gannet.stats.ox.ac.uk>
Prof Brian Ripley ?????:
> On Sun, 19 Feb 2006, Evgeniy Kachalin wrote:
>
>> Hello, dear R users.
>>
>> What is the easiest and the most visualli understandable way to analize
>> dependency of numerical variable on two factors?
>
> interaction.plot() is a good start.
>
>> Is the
>> boxplot(y~f1+f2) the good way? It seems that this formula does not work.
>
> No, nor is it documented to: the help page is there to help you. You
> need a single factor as the grouping, so make one via an interaction.
> boxplot(y ~ f1:f2) should work. E.g.
>
> library(MASS)
> boxplot(FL ~ sex:sp, data=crabs)
Does not work:
???????????? ?? if (any(out[nna])) stats[c(1, 5)] <- range(x[!out], na.rm =
TRUE) :
?????????????????????? ????????????????, ?? ?????????? TRUE/FALSE
????????????????: Warning messages:
1: + not meaningful for factors in: Ops.factor(x[floor(d)], x[ceiling(d)])
2: < not meaningful for factors in: Ops.factor(x, (stats[2] - coef * iqr))
3: > not meaningful for factors in: Ops.factor(x, (stats[4] + coef * iqr))
Hm...
> Another idea is to use lattice's bwplot. E.g.
>
> library(lattice)
> bwplot(FL ~ sex | sp, data=crabs)
>
>
That's not the point. The scales may differ significantly, also this is
not conviniet for many factors.