R help - Mar 2011 - Generating repeated measures data

How would one generate data to be used in a simulation of a repeated measures
ANOVA given a known (1) within-person correlation with known (2) mean and SD of
data obtained at each of three times of observation?
Thanks,
John
John Sorkin
Chief Biostatistics and Informatics
Univ. of Maryland School of Medicine
Division of Gerontology and Geriatric Medicine
JSorkin at grecc.umaryland.edu 
Confidentiality Statement:
This email message, including any attachments, is for th...{{dropped:6}}

Hi John,

This is not advanced enough for serious applications, but if you are
just looking for a simple way to make example data, you could do
something like:

gencor <- function(order, mu, sd, r) {
  X <- matrix(rnorm(prod(order), mean = mu, sd = sd), nrow = order[1],
ncol = order[2])
  R <- matrix(r, nrow = order[2], ncol = order[2])
  diag(R) <- 1
  X %*% chol(R)
}

cor(X <- gencor(c(100, 3), 10, 2, .5))

HTH,

Josh

On Fri, Mar 18, 2011 at 5:37 PM, John Sorkin
<jsorkin at grecc.umaryland.edu> wrote:
> How would one generate data to be used in a simulation of a repeated
measures ANOVA given a known (1) within-person correlation with known (2) mean
and SD of data obtained at each of three times of observation?
> Thanks,
> John
> John Sorkin
> Chief Biostatistics and Informatics
> Univ. of Maryland School of Medicine
> Division of Gerontology and Geriatric Medicine
> JSorkin at grecc.umaryland.edu
> Confidentiality Statement:
> This email message, including any attachments, is for ...{{dropped:17}}

At 00:37 19/03/2011, John Sorkin wrote:
>How would one generate data to be used in a simulation of a repeated 
>measures ANOVA given a known (1) within-person correlation with 
>known (2) mean and SD of data obtained at each of three times of
observation?

You do not say which distribution you want them to have but for 
normal the best choice seems to be to load MASS and use rmvtnorm.

Generating multivariate datasets from distributions
other than the normal
can be more challenging.
In \R contributed packages provide the following among others:
\begin{description}
\item[\pkg{corcount}]
Poisson, negative binomial, zero--inflated versions
\item[\pkg{binarySimCLF}]
Generates binary variables, see \citet{qaqish03}
for details
\item[\pkg{bindat}]
Generates binary by thresholding normal
\item[\pkg{mvtBinaryEP}]
Binary
\item[\pkg{sn}]
Skew normal and skew $t$
\end{description}

Apologies for all the LaTeX but I fear if I try to edit it I will 
delete something crucial.

>Thanks,
>John
>John Sorkin
>Chief Biostatistics and Informatics
>Univ. of Maryland School of Medicine
>Division of Gerontology and Geriatric Medicine
>JSorkin at grecc.umaryland.edu
>Confidentiality Statement:
>This email message, including any attachments, is for t...{{dropped:6}}

hi
i checked the package "corcounts", but it seems to be used for
generating
high dimensional correlated count data.  is it also proper for generating
data with N=50  for instance?

thanks

--
View this message in context:
http://r.789695.n4.nabble.com/Generating-repeated-measures-data-tp3388807p4409373.html
Sent from the R help mailing list archive at Nabble.com.