SPSSX Discussion

Significance testing?

Classic

List

Threaded

5 messages Options

Michael C. Morrison

Significance testing?

I'm conducting a multiple regression analysis with all 50 States as the
dependent variable. Looking at the output and out of habit I immediately
looked at the F-ratio and p-value. Then I thought wait a second I just did
an analysis of the entire population, not a sample, so the F-ratio and
regression coefficient tests of significance have no meaning. Correct? Any
articles on this topic?

In advance, thanks.

Mike

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

SR Millis-3

Re: Significance testing?

Mike,
I'm curious: how are you doing this? The US states variable is categorical in scaling. Multiple regression assumes a continuous response variable.

Scott R Millis, PhD, ABPP (CN,CL,RP), CStat, CSci
Professor & Director of Research
Dept of Physical Medicine & Rehabilitation
Dept of Emergency Medicine
Wayne State University School of Medicine
261 Mack Blvd
Detroit, MI 48201
Email: [hidden email]
Tel: 313-993-8085
Fax: 313-966-7682

--- On Mon, 6/22/09, Mike <[hidden email]> wrote:

> From: Mike <[hidden email]>
> Subject: Significance testing?
> To: [hidden email]
> Date: Monday, June 22, 2009, 1:53 PM
> I'm conducting a multiple regression
> analysis with all 50 States as the
> dependent variable. Looking at the output and out of habit
> I immediately
> looked at the F-ratio and p-value. Then I thought wait a
> second I just did
> an analysis of the entire population, not a sample, so the
> F-ratio and
> regression coefficient tests of significance have no
> meaning. Correct? Any
> articles on this topic?
>
> In advance, thanks.
>
> Mike
>
> =====================
> 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
>

Hector Maletta

Re: Significance testing?

In reply to this post by Michael C. Morrison

You have not taken a sample of States out of the population of all 50
states, but you have taken a sample of measurements (50 measurements) out of
a population of potentially many measurements of the variables in different
states. Variables in statistical procedures are often regarded as random
variables, the value of which may vary randomly from one measurement to the
next.
For instance, if you are dealing with State attributes such as winter
minimum temperature or percentage of people over 65, you may have taken the
same measurements the year before, or the month before, or whatever. Your
particular data set might be seen as just one sample set out of many
possible sets of 50 measurements of the variables. Sample values might
randomly (albeit slightly) vary across such replications.
Under this approach, each particular State is but a "case" of a general
class of homogeneous objects called "States", or "States at time t". That is
the reason why you put them all in a statistical table and subjected them to
a regression analysis. If they are not members of a homogeneous class in any
meaningful sense, they would not be fit for such an analysis and each should
be treated as a singular object, entire of itself like the Big Bang or the
French Revolution.
So, depending on the theoretical approach you are implicitly using to
examine the 50 States of the Union, significance testing may make sense,
even if you covered the entire list of States. Or it may not.
Oops, this is perhaps getting too dense, too philosophical, or too
epistemological. Sorry.
Hector

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Mike
Sent: 22 June 2009 14:53
To: [hidden email]
Subject: Significance testing?

I'm conducting a multiple regression analysis with all 50 States as the
dependent variable. Looking at the output and out of habit I immediately
looked at the F-ratio and p-value. Then I thought wait a second I just did
an analysis of the entire population, not a sample, so the F-ratio and
regression coefficient tests of significance have no meaning. Correct? Any
articles on this topic?

In advance, thanks.

Mike

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

David Greenberg

Re: Significance testing?

In reply to this post by Michael C. Morrison

Usually the states would be the unit of analysis in a regression that predicts a characteristic of the state as a dependent variable, with other state characteristics as predictors. Once Hawaii and Alaska became states, the status of each territory as a state has been fixed, unchanged, and couldn't be the dependent variable in a regression analysis. David Greenberg, Sociology Department, New York University

----- Original Message -----
From: Mike <[hidden email]>
Date: Monday, June 22, 2009 3:36 pm
Subject: Significance testing?
To: [hidden email]

> I'm conducting a multiple regression analysis with all 50 States as the
> dependent variable. Looking at the output and out of habit I immediately
> looked at the F-ratio and p-value. Then I thought wait a second I
> just did
> an analysis of the entire population, not a sample, so the F-ratio and
> regression coefficient tests of significance have no meaning.
> Correct? Any
> articles on this topic?
>
> In advance, thanks.
>
> Mike
>
> =====================
> 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
>

Hector Maletta

Re: Significance testing?

When I answered Mike's question I thought, as David and others, that the
question was probably wrongly phrased, since it says that the analysis had
"all 50 States as the dependent variable". That wouldn't make sense, so I
interpreted it as meaning that the data set was comprised of the 50 states,
and the dependent variable was some characteristic of the States. If it was
not so, then the question needs rephrasing.
Hector
-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
David Greenberg
Sent: 22 June 2009 19:03
To: [hidden email]
Subject: Re: Significance testing?

Usually the states would be the unit of analysis in a regression that
predicts a characteristic of the state as a dependent variable, with other
state characteristics as predictors. Once Hawaii and Alaska became states,
the status of each territory as a state has been fixed, unchanged, and
couldn't be the dependent variable in a regression analysis. David
Greenberg, Sociology Department, New York University

----- Original Message -----
From: Mike <[hidden email]>
Date: Monday, June 22, 2009 3:36 pm
Subject: Significance testing?
To: [hidden email]

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