Transformation with large sample

[Maria Sapouna]

Dear colleagues,

I would like to perform a regression on a sample of 1758 participants but
my dependent variables are not normally distributed. I have read that with
large samples transformation is not necessary. Is this right?

Thank you for all your help.

Best wishes

Mary

=====================
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

[Alexander J. Shackman-2]

No, it is not. The assumptions of the general linear model are independent
of sample size. You may be thinking of the notion that larger samples simply
tend to have more normal distributions (assuming the parent population is
normal)...cf. http://en.wikipedia.org/wiki/Central_limit_theorem.
How did you determine that your sample is non-normal? With such a large
size, even minor deviations will be significantly non-normal using the usual
tests (ie, they're overpowered for testing the normality). Depending on the
nature of the variable (ie, whether the pop is likely to be non-normal, as
with reaction times) and magnitude of the non-normality you may be able to
use the data as-is. Otherwise, transform in the usual way.

See also
http://psyphz.psych.wisc.edu/~shackman/mediation_moderation_resources.htm#DataDiagnosis
http://psyphz.psych.wisc.edu/~shackman/mediation_moderation_resources.htm#DataTransforms


HTH, Alex
On Jan 6, 2008 9:05 AM, M Sapona <[hidden email]> wrote:

> Dear colleagues,
>
> I would like to perform a regression on a sample of 1758 participants but
> my dependent variables are not normally distributed. I have read that with
> large samples transformation is not necessary. Is this right?
>
> Thank you for all your help.
>
> Best wishes
>
> Mary
>
> =====================
> 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
>

--
Alexander J. Shackman
Laboratory for Affective Neuroscience
Waisman Laboratory for Brain Imaging & Behavior
University of Wisconsin-Madison
1202 West Johnson Street
Madison, Wisconsin 53706

Telephone: [phone]
FAX: [phone]
EMAIL: [hidden email]
http://psyphz.psych.wisc.edu/~shackman
Calendar {still under construction}:
http://www.google.com/calendar/embed?src=ajshackman%40gmail.com

=====================
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

[Art Kendall-2]

The assumption you want to check is not whether the raw data are
normally distributed, but whether or not the distribution of the
*residuals* is severely discrepant from normal.

Art Kendall
Social Research Consultants.

M Sapona wrote:

> Dear colleagues,
>
> I would like to perform a regression on a sample of 1758 participants but
> my dependent variables are not normally distributed. I have read that with
> large samples transformation is not necessary. Is this right?
>
> Thank you for all your help.
>
> Best wishes
>
> Mary
>
> =====================
> 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

[Alexander J. Shackman-2]

agreed

On Jan 6, 2008 10:08 AM, Art Kendall <[hidden email]> wrote:

> The assumption you want to check is not whether the raw data are
> normally distributed, but whether or not the distribution of the
> *residuals* is severely discrepant from normal.
>
> Art Kendall
> Social Research Consultants.
>
> M Sapona wrote:
> > Dear colleagues,
> >
> > I would like to perform a regression on a sample of 1758 participants
> but
> > my dependent variables are not normally distributed. I have read that
> with
> > large samples transformation is not necessary. Is this right?
> >
> > Thank you for all your help.
> >
> > Best wishes
> >
> > Mary
> >
> > =====================
> > 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
>

--
Alexander J. Shackman
Laboratory for Affective Neuroscience
Waisman Laboratory for Brain Imaging & Behavior
University of Wisconsin-Madison
1202 West Johnson Street
Madison, Wisconsin 53706

Telephone: [phone]
FAX: [phone]
EMAIL: [hidden email]
http://psyphz.psych.wisc.edu/~shackman
Calendar {still under construction}:
http://www.google.com/calendar/embed?src=ajshackman%40gmail.com

=====================
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