R help - Feb 2007 - How-To construct a cov list to use a covariance matrix in factanal?

Hi,

I have a set of covariance matrices but not the original data. I want to carry
out some exploratory factor analysis. So, I am trying to construct a covariance
matrix list as the input for factanal. I can construct a list which includes the
cov, the centers, and the n.obs. But it doesn't work. I get an error that
says "Error in sqrt(diag(cv)) : Non-numeric argument to mathematical
function". So, obviously I am doing something wrong.

Two questions occur. Can someone either tell me how to construct a proper covmat
list object or point me to a description of how to do it? The other question is
whether it is possible to simply use the covariance matrix as the argument for
covmat in factanal? The description implies that it is but I really have no idea
of how to do this. I have tried simply making covmat the covariance matrix but
it doesn't wotk. I just get the message "'covmat' is not a
valid covariance list"

Anyway, thanks for any thoughts you might have on this.

Alistair Campbell
-- 
Dr Alistair Campbell, PhD
Senior Lecturer in Clinical Psychology
School of Psychology
James Cook University
Townsville Queensland Australia

Ph: +61 7 47816879

The help page says

   covmat: A covariance matrix, or a covariance list as returned by
           'cov.wt'.  Of course, correlation matrices are covariance
           matrices.

and there is an example of a covariance list (ability.cov).
> factanal(factors = 2, covmat = ability.cov)
> factanal(factors = 2, covmat = ability.cov$cov)
both work.  See
> PLEASE do read the posting guide
http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.
as we have no idea what you tried.

On Tue, 6 Feb 2007, Alistair Campbell wrote:
> Hi,
>
> I have a set of covariance matrices but not the original data. I want to 
> carry out some exploratory factor analysis. So, I am trying to construct 
> a covariance matrix list as the input for factanal. I can construct a 
> list which includes the cov, the centers, and the n.obs. But it doesn't
> work. I get an error that says "Error in sqrt(diag(cv)) : Non-numeric 
> argument to mathematical function". So, obviously I am doing something
> wrong.
>
> Two questions occur. Can someone either tell me how to construct a 
> proper covmat list object or point me to a description of how to do it? 
> The other question is whether it is possible to simply use the 
> covariance matrix as the argument for covmat in factanal? The 
> description implies that it is but I really have no idea of how to do 
> this. I have tried simply making covmat the covariance matrix but it 
> doesn't wotk. I just get the message "'covmat' is not a
valid covariance
> list"
>
> Anyway, thanks for any thoughts you might have on this.
>
> Alistair Campbell
>
-- 
Brian D. Ripley,                  ripley at stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595

Thanks for that Brian,

I have worked through the examples. They work because the covmat were produced
by the cov.wt which provides output as a list object. I am trying to construct
my own list object to use as the covmat. There are no obvious instructions on
how to do this.

So, here is what I have done so far.

I reconstructed the covariance matrix in the example and created a dataframe:

 > testmatrix
  general picture  blocks   maze reading   vocab
1  24.641   5.991  33.520  6.023  20.755  29.701
2   5.991   6.700  18.137  1.782   4.936   7.204
3  33.520  18.137 149.831 19.424  31.430  50.753
4   6.023   1.782  19.424 12.711   4.757   9.075
5  20.755   4.936  31.430  4.757  52.604  66.762
6  29.701   7.204  50.753  9.075  66.762 135.292

and then used this to construct a list object like the output from the example;
> tstcov<- list(cov=testmatrix, center=c(0,0,0,0,0), n.obs=112)
I tested to see whether my list object looked like the examples
> tstcov
$cov
  general picture  blocks   maze reading   vocab
1  24.641   5.991  33.520  6.023  20.755  29.701
2   5.991   6.700  18.137  1.782   4.936   7.204
3  33.520  18.137 149.831 19.424  31.430  50.753
4   6.023   1.782  19.424 12.711   4.757   9.075
5  20.755   4.936  31.430  4.757  52.604  66.762
6  29.701   7.204  50.753  9.075  66.762 135.292

$centers
[1] 0 0 0 0 0

$n.obs
[1] 112

It looks the same. So I then used this list as the argument in factanal and get
the error message.
> factanal(factors=2, covmat=tstcov, rotation="varimax")
Error in sqrt(diag(cv)) : Non-numeric argument to mathematical function

I know that what you see of a list is not necessarily all that is there. So, I
figure I am missing some part of the object that makes this list suitable for
use by factanal.

So, I hope this is enough detail. Any thoughts would be appreciated.
-- 
Dr Alistair Campbell, PhD
Senior Lecturer in Clinical Psychology
School of Psychology
James Cook University
Townsville Queensland Australia

Ph: +61 7 47816879

Alistair,

>
>I have worked through the examples. They work because the covmat 
>were produced by the cov.wt which provides output as a list object. 
>I am trying to construct my own list object to use as the covmat. 
>There are no obvious instructions on how to do this.
>
>So, here is what I have done so far.
>
>I reconstructed the covariance matrix in the example and created a
dataframe:
>
>  > testmatrix
>   general picture  blocks   maze reading   vocab
>1  24.641   5.991  33.520  6.023  20.755  29.701
>2   5.991   6.700  18.137  1.782   4.936   7.204
>3  33.520  18.137 149.831 19.424  31.430  50.753
>4   6.023   1.782  19.424 12.711   4.757   9.075
>5  20.755   4.936  31.430  4.757  52.604  66.762
>6  29.701   7.204  50.753  9.075  66.762 135.292
>
>and then used this to construct a list object like the output from 
>the example;
>
>>  tstcov<- list(cov=testmatrix, center=c(0,0,0,0,0), n.obs=112)
>
>I tested to see whether my list object looked like the examples
>
>>  tstcov
>$cov
>   general picture  blocks   maze reading   vocab
>1  24.641   5.991  33.520  6.023  20.755  29.701
>2   5.991   6.700  18.137  1.782   4.936   7.204
>3  33.520  18.137 149.831 19.424  31.430  50.753
>4   6.023   1.782  19.424 12.711   4.757   9.075
>5  20.755   4.936  31.430  4.757  52.604  66.762
>6  29.701   7.204  50.753  9.075  66.762 135.292
>
>$centers
>[1] 0 0 0 0 0
>
>$n.obs
>[1] 112
>
>It looks the same. So I then used this list as the argument in 
>factanal and get the error message.
>
>>  factanal(factors=2, covmat=tstcov, rotation="varimax")
>Error in sqrt(diag(cv)) : Non-numeric argument to mathematical function
>
>I know that what you see of a list is not necessarily all that is 
>there. So, I figure I am missing some part of the object that makes 
>this list suitable for use by factanal.
>
So, I hope this is enough detail. Any thoughts would be
appreciated.
>--
>Dr Alistair Campbell, PhD
>Senior Lecturer in Clinical Psychology
>School of Psychology
>James Cook University
Townsville Queensland Australia



The call to factanal requires you to specify that you have a 
covariance matrix (covmat).  Using your matrix and specifying two 
factors:


test.df
   general picture  blocks   maze reading   vocab
1  24.641   5.991  33.520  6.023  20.755  29.701
2   5.991   6.700  18.137  1.782   4.936   7.204
3  33.520  18.137 149.831 19.424  31.430  50.753
4   6.023   1.782  19.424 12.711   4.757   9.075
5  20.755   4.936  31.430  4.757  52.604  66.762
6  29.701   7.204  50.753  9.075  66.762 135.292
>  test.mat <- as.matrix(test.df)
>  tf <- factanal(factors=2,covmat=test.mat)
>  tf
Call:
factanal(factors = 2, covmat = test.mat)

Uniquenesses:
general picture  blocks    maze reading   vocab
   0.455   0.589   0.218   0.769   0.052   0.334

Loadings:
   Factor1 Factor2
1 0.499   0.543 
2 0.156   0.622 
3 0.206   0.860 
4 0.109   0.468 
5 0.956   0.182 
6 0.785   0.225 

                Factor1 Factor2
SS loadings      1.858   1.724
Proportion Var   0.310   0.287
Cumulative Var   0.310   0.597

The degrees of freedom for the model is 4 and the fit was 0.0572

Bill
>
-- 
William Revelle		http://pmc.psych.northwestern.edu/revelle.html   
Professor			http://personality-project.org/personality.html
Department of Psychology       http://www.wcas.northwestern.edu/psych/
Northwestern University	http://www.northwestern.edu/