R help - Sep 2003 - what does this error mean?

Dear R-users:
I am runing R 1.6.2 with Windows XP. I try to calculate Pearson correlation 
and Spearman correlation of any pairwise columns of 8000 x 80 data matrix with 
missing values and randomize the matix 1000 times and calculate this two 
correlations again. The code bellow for Pearson is working fine but for 
Spearman got the error bellow for randomized data matrix and R console is stop 
working: 

cor (raw2[,i], raw2[,j], use="complete.obs")
cor.test (raw2[,i], raw2[,j],  method="spearman", na.action =
"na.omit")
$estimate

Error in cor.test.default(raw2[, i], raw2[, j], method = "spearman",
na.action
= "na.omit") : 
        not enough finite observations
In addition: There were 50 or more warnings (use warnings() to see the first 
50)

What does the Error mean? How to fix it?
Any suggestion are highly appreciated?
Josh

cor.test does not have an na.action argument unless used with the formula
interface, when it is better to use na.action=na.omit, not "na.omit".

The error message occurs if less than 2 rows are completely non-missing.
So take a closer look at your data.

BTW, I would compare the different methods of cor.test, not cor.test with 
cor, to check I really had this right.


On Sat, 13 Sep 2003 szhan at uoguelph.ca wrote:
> Dear R-users:
> I am runing R 1.6.2 with Windows XP. I try to calculate Pearson correlation
That's long obselete, as you know.
> and Spearman correlation of any pairwise columns of 8000 x 80 data
> matrix with missing values and randomize the matix 1000 times and
> calculate this two correlations again. The code bellow for Pearson is
> working fine but for Spearman got the error bellow for randomized data
> matrix and R console is stop working:
And what does that last phrase mean?  (And `bellow' is not something you
should do on a public list.)
> cor (raw2[,i], raw2[,j], use="complete.obs")
> cor.test (raw2[,i], raw2[,j],  method="spearman", na.action =
"na.omit")
> $estimate
> 
> Error in cor.test.default(raw2[, i], raw2[, j], method =
"spearman", na.action
> = "na.omit") : 
>         not enough finite observations
> In addition: There were 50 or more warnings (use warnings() to see the
first
> 50)
> 
> What does the Error mean? How to fix it?
> Any suggestion are highly appreciated?
> Josh

-- 
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

Josh  -

See the example "run a simulation" in  help("try").

This won't be able to tell you the Spearman correlation
when there are fewer than two pairs of non-missing values,
but it will allow the loop to keep running.  Caution: the
return value from  try()  will be a string with the error
message, when  cor.test()  fails, so whatever variable
you accumulate the simulation results in must be able
to tolerate string values.  That suggests using a list.

-  tom blackwell  -  u michigan medical school  -  ann arbor  -

On Sat, 13 Sep 2003 szhan at uoguelph.ca wrote:
> Dear R-users:
> I am runing R 1.6.2 with Windows XP. I try to calculate Pearson correlation
> and Spearman correlation of any pairwise columns of 8000 x 80 data matrix
with
> missing values and randomize the matix 1000 times and calculate this two
> correlations again. The code bellow for Pearson is working fine but for
> Spearman got the error bellow for randomized data matrix and R console is
stop
> working:
>
> cor (raw2[,i], raw2[,j], use="complete.obs")
> cor.test (raw2[,i], raw2[,j],  method="spearman", na.action =
"na.omit")
> $estimate
>
> Error in cor.test.default(raw2[, i], raw2[, j], method =
"spearman",
> na.action = "na.omit") :
>         not enough finite observations
> In addition: There were 50 or more warnings (use warnings() to see
> the first  50)
>
> What does the Error mean? How to fix it?
> Any suggestion are highly appreciated?
> Josh