comparing correlations among two groups

[Nomi-2]

Hello all,
I've checked the correlation between 2 variables in a group of patients
with a disease and in a group of healthy controls. There were significant
weak correlations in both groups. Is there anyway to compare the R values
of the 2 groups so that I know whether there is a significant difference
in the correlation between the cases and the controls?
Thanks!
Nomi

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

[Julius Sim]

Yes, there is - but you first have to transform the correlations to a
linear scale using Fisher's method:

z = .5[ln(1+r) - ln(1-r)]

You can then get a standard normal deviate, and thereby a corresponding p
value using tables, through this formula:

z = (z1-z2)/sqrt[1/(n1-3) + 1/(n2-3)]

where z is the standard normal deviate for the test and z1 and z2 are the
Fisher-transformed correlations for the two samples.

There might be a stats program that will do this for you!

Julius

> Hello all,
> I've checked the correlation between 2 variables in a group of patients
> with a disease and in a group of healthy controls. There were significant
> weak correlations in both groups. Is there anyway to compare the R values
> of the 2 groups so that I know whether there is a significant difference
> in the correlation between the cases and the controls?
> Thanks!
> Nomi
>
> =====================
> To manage your subscription to SPSSX-L, send a message to
> [hidden email] (not to SPSSX-L), with no body text except the
> command. To leave the list, send the command
> SIGNOFF SPSSX-L
> For a list of commands to manage subscriptions, send the command
> INFO REFCARD
>

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

[Marta Garcia-Granero]

Julius Sim escribió:

> Yes, there is - but you first have to transform the correlations to a
> linear scale using Fisher's method:
>
> z = .5[ln(1+r) - ln(1-r)]
>
> You can then get a standard normal deviate, and thereby a corresponding p
> value using tables, through this formula:
>
> z = (z1-z2)/sqrt[1/(n1-3) + 1/(n2-3)]
>
> where z is the standard normal deviate for the test and z1 and z2 are the
> Fisher-transformed correlations for the two samples.
>
> There might be a stats program that will do this for you!
>
>
>

Yes, a freeware statistics calculator called Simcalc (you can download
it with SIMSTAT demo, from Provalis Research page)

HTH,
Marta García-Granero

>> I've checked the correlation between 2 variables in a group of patients
>> with a disease and in a group of healthy controls. There were significant
>> weak correlations in both groups. Is there anyway to compare the R values
>> of the 2 groups so that I know whether there is a significant difference
>> in the correlation between the cases and the controls?
>> Thanks!
>> Nomi
>>

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

[Johnny Amora]

Multiple group comparisons of the following are very
easy using AMOS: Correlation coefficents, Regression
coefficients, path coefficients, factor loadings,
coefficients of structural equation models, among
others.

Cheers,

Johnny




--- Julius Sim <[hidden email]> wrote:

> Yes, there is - but you first have to transform the
> correlations to a
> linear scale using Fisher's method:
>
> z = .5[ln(1+r) - ln(1-r)]
>
> You can then get a standard normal deviate, and
> thereby a corresponding p
> value using tables, through this formula:
>
> z = (z1-z2)/sqrt[1/(n1-3) + 1/(n2-3)]
>
> where z is the standard normal deviate for the test
> and z1 and z2 are the
> Fisher-transformed correlations for the two samples.
>
> There might be a stats program that will do this for
> you!
>
> Julius
>
>
>
> > Hello all,
> > I've checked the correlation between 2 variables
> in a group of patients
> > with a disease and in a group of healthy controls.
> There were significant
> > weak correlations in both groups. Is there anyway
> to compare the R values
> > of the 2 groups so that I know whether there is a
> significant difference
> > in the correlation between the cases and the
> controls?
> > Thanks!
> > Nomi
> >
> > =====================
> > To manage your subscription to SPSSX-L, send a
> message to
> > [hidden email] (not to SPSSX-L), with
> no body text except the
> > command. To leave the list, send the command
> > SIGNOFF SPSSX-L
> > For a list of commands to manage subscriptions,
> send the command
> > INFO REFCARD
> >
>
> =====================
> To manage your subscription to SPSSX-L, send a
> message to
> [hidden email] (not to SPSSX-L), with no
> body text except the
> command. To leave the list, send the command
> SIGNOFF SPSSX-L
> For a list of commands to manage subscriptions, send
> the command
> INFO REFCARD
>

      ____________________________________________________________________________________
Looking for last minute shopping deals?
Find them fast with Yahoo! Search.  http://tools.search.yahoo.com/newsearch/category.php?category=shopping

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD