SPSSX Discussion

Geometric mean of a variable

Classic

List

Threaded

16 messages Options

drfg2008

Geometric mean of a variable

SPSS 20

How can I compute a geometric mean of a variable ?
(The problem is, that the variable only consists of positive numbers LE 1 and N of numbers is large -> 38.000)

Example: What is the geometric mean of VAR001 ?

input program.
loop a =1 to 38*10**3 by 1.
end case.
end loop.
end file.
end input program.
EXECUTE.
COMPUTE VAR001 =RV.UNIFORM(0,1).
EXECUTE.

Thanks.

Dr. Frank Gaeth

Art Kendall

Re: Geometric mean of a variable

input program.
loop a =1 to 38*10**3 by 1.
COMPUTE VAR001 =RV.UNIFORM(0,1).
end case.
end loop.
end file.
end input program.
MEANS TABLES=VAR001
/CELLS MEAN COUNT STDDEV GEOMETRIC SUM.


VAR001
Mean    N    Std. Deviation    Geometric Mean    Sum
.50    38000    .289    .37    18996

Resources    Processor Time    00:00:00.08
    Elapsed Time    00:00:00.14

Art Kendall
Social Research Consultants

On 12/9/2012 3:26 PM, drfg2008 wrote:

input program.
loop a =1 to 38*10**3 by 1.
end case.
end loop.
end file.
end input program.
EXECUTE.
COMPUTE VAR001 =RV.UNIFORM(0,1).
EXECUTE.

===================== 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
Social Research Consultants

Mike

Re: Geometric mean of a variable

In reply to this post by drfg2008

Someone may be able to provide the relevant code but this is what you do:

(1) Create a new variable logvar001=log(var001).

(2) Create the sum of logvar001 (probably through aggregate and add to
active dataset).

(3) Compute the antilog of the sum of logvar001.

compute alogsum=10**sumlogvar001.

Here are examples of how to compute the Geometric mean by hand or whatever:
http://books.google.com/books?id=3C9LzilnF6YC&pg=PA83&dq=%22calculation+of+geometric+mean%22&hl=en&sa=X&ei=0_jEULSPN8i80QH28wE&ved=0CDMQ6AEwAQ#v=onepage&q=%22calculation%20of%20geometric%20mean%22&f=false

And here is an example of how to get an antilog in SPSS:
http://www-01.ibm.com/support/docview.wss?uid=swg21480979

I'm sure that if I'm error someone will correct me.

-Mike Palij
New York University
[hidden email]

----- Original Message -----
From: "drfg2008" <[hidden email]>
To: <[hidden email]>
Sent: Sunday, December 09, 2012 3:26 PM
Subject: Geometric mean of a variable

> SPSS 20
>
> How can I compute a geometric mean of a variable ?
> (The problem is, that the variable only consists of positive numbers LE 1
> and N of numbers is large -> 38.000)
>
> Example: What is the geometric mean of VAR001 ?
>
> input program.
> loop a =1 to 38*10**3 by 1.
> end case.
> end loop.
> end file.
> end input program.
> EXECUTE.
> COMPUTE VAR001 =RV.UNIFORM(0,1).
> EXECUTE.
>
> Thanks.

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

Mike

Re: Geometric mean of a variable

In reply to this post by Art Kendall

Now that is interesting. I double-checked the v20 manual and

see that both the means and summarize procedure provide the

geometric mean. I only have two questions:

(1) When did this become available?

and

(2) Why aren't these provided in frequencies, descriptives,

and examine/explore?

-Mike Palij

New York University

[hidden email]

----- Original Message -----

From: [hidden email]

To: [hidden email]

Sent: Sunday, December 09, 2012 3:57 PM

Subject: Re: Geometric mean of a variable
input program.
loop a =1 to 38*10**3 by 1.
COMPUTE VAR001 =RV.UNIFORM(0,1).
end case.
end loop.
end file.
end input program.
MEANS TABLES=VAR001
/CELLS MEAN COUNT STDDEV GEOMETRIC SUM.


VAR001
Mean    N    Std. Deviation    Geometric Mean    Sum
.50    38000    .289    .37    18996

Resources    Processor Time    00:00:00.08
    Elapsed Time    00:00:00.14
Art Kendall
Social Research Consultants
On 12/9/2012 3:26 PM, drfg2008 wrote:
input program.
loop a =1 to 38*10**3 by 1.
end case.
end loop.
end file.
end input program.
EXECUTE.
COMPUTE VAR001 =RV.UNIFORM(0,1).
EXECUTE.
===================== 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

Jon K Peck

Re: Geometric mean of a variable

The oldest version of SPSS I have installed is 15, and the geometric mean is available there. Since there is no update notice for it in the command history, it is probably present at least back to V13 when command histories were introduced.

But why would you expect every descriptive procedure to offer all the same statistics, especially ones that are little used?

Jon Peck (no "h") aka Kim
Senior Software Engineer, IBM
[hidden email]
new phone: 720-342-5621

From: Mike Palij <[hidden email]>
To: [hidden email],
Date: 12/09/2012 03:15 PM
Subject: Re: [SPSSX-L] Geometric mean of a variable
Sent by: "SPSSX(r) Discussion" <[hidden email]>

Now that is interesting. I double-checked the v20 manual and
see that both the means and summarize procedure provide the
geometric mean. I only have two questions:

(1) When did this become available?

and

(2) Why aren't these provided in frequencies, descriptives,
and examine/explore?

-Mike Palij
New York University
mp26@...
----- Original Message -----
From: Art Kendall
To: [hidden email]
Sent: Sunday, December 09, 2012 3:57 PM
Subject: Re: Geometric mean of a variable

input program.
loop a =1 to 38*10**3 by 1.
COMPUTE VAR001 =RV.UNIFORM(0,1).
end case.
end loop.
end file.
end input program.
MEANS TABLES=VAR001
/CELLS MEAN COUNT STDDEV GEOMETRIC SUM.

VAR001
Mean N Std. Deviation Geometric Mean Sum
.50 38000 .289 .37 18996

Resources Processor Time 00:00:00.08
Elapsed Time 00:00:00.14

Art Kendall Social Research Consultants
On 12/9/2012 3:26 PM, drfg2008 wrote:
input program. loop a =1 to 38*10**3 by 1. end case. end loop. end file. end input program. EXECUTE. COMPUTE VAR001 =RV.UNIFORM(0,1). EXECUTE.

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

Ryan

Re: Geometric mean of a variable

In reply to this post by drfg2008

By fitting an intercept-only general linear model on the natural log of the variable and then exponentiating the estimated intercept and confidence limits will provide both the geometric mean and the 95% confidence interval, respectively:

input program.

loop a =1 to 38*10**3 by 1.

end case.

end loop.

end file.

end input program.

EXECUTE.

COMPUTE VAR001 =RV.UNIFORM(0,1).

EXECUTE.

compute natural_log_VAR001=ln(VAR001).

execute.

UNIANOVA natural_log_VAR001

/METHOD=SSTYPE(3)

/INTERCEPT=INCLUDE

/PRINT=PARAMETER

/CRITERIA=ALPHA(.05).

compute geometric_mean = exp(-.9951088706552569).

compute geometric_mean_LL = exp(-1.005089860177377).

compute geometric_mean_UL = exp(-.9851278811331369).

execute.

Ryan

On Sun, Dec 9, 2012 at 3:26 PM, drfg2008 <[hidden email]> wrote:

> SPSS 20

> How can I compute a geometric mean of a variable ?

> (The problem is, that the variable only consists of positive numbers LE 1

> and N of numbers is large -> 38.000)

> Example: What is the geometric mean of VAR001 ?

> input program.

> loop a =1 to 38*10**3 by 1.

> end case.

> end loop.

> end file.

> end input program.

> EXECUTE.

> COMPUTE VAR001 =RV.UNIFORM(0,1).

> EXECUTE.

> Thanks.

> -----

> Dr. Frank Gaeth

> FU-Berlin

> --

> View this message in context: http://spssx-discussion.1045642.n5.nabble.com/Geometric-mean-of-a-variable-tp5716795.html

> Sent from the SPSSX Discussion mailing list archive at Nabble.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

David Marso

Re: Geometric mean of a variable

Administrator

In reply to this post by Mike

Close:
But
1. SPSS doesn't support LOG.
Use either LN or LG10.
LG10(numexpr)
Returns the base-10 logarithm of numexpr, which must be numeric and greater than 0.
LN(numexpr)
Returns the base-e logarithm of numexpr, which must be numeric and greater than 0.
2. You left out a final step (divide by 10).
---
So .. relevant code (untested)**.
--
COMPUTE newvar=LG10(oldvar).
AGGREGATE OUTFILE * MODE=ADDVARIABLES / Snewvar=SUM(newvar) / N=N(newvar).
COMPUTE GeoMean=10^(SNewvar/N).
---

Mike Palij wrote

Someone may be able to provide the relevant code but this is what you do:

(1) Create a new variable logvar001=log(var001).

(2) Create the sum of logvar001 (probably through aggregate and add to
active dataset).

(3) Compute the antilog of the sum of logvar001.

compute alogsum=10**sumlogvar001.

Here are examples of how to compute the Geometric mean by hand or whatever:
http://books.google.com/books?id=3C9LzilnF6YC&pg=PA83&dq=%22calculation+of+geometric+mean%22&hl=en&sa=X&ei=0_jEULSPN8i80QH28wE&ved=0CDMQ6AEwAQ#v=onepage&q=%22calculation%20of%20geometric%20mean%22&f=false

And here is an example of how to get an antilog in SPSS:
http://www-01.ibm.com/support/docview.wss?uid=swg21480979

I'm sure that if I'm error someone will correct me.

-Mike Palij
New York University
[hidden email]

----- Original Message -----
From: "drfg2008" <[hidden email]>
To: <[hidden email]>
Sent: Sunday, December 09, 2012 3:26 PM
Subject: Geometric mean of a variable

> SPSS 20
>
> How can I compute a geometric mean of a variable ?
> (The problem is, that the variable only consists of positive numbers LE 1
> and N of numbers is large -> 38.000)
>
> Example: What is the geometric mean of VAR001 ?
>
> input program.
> loop a =1 to 38*10**3 by 1.
> end case.
> end loop.
> end file.
> end input program.
> EXECUTE.
> COMPUTE VAR001 =RV.UNIFORM(0,1).
> EXECUTE.
>
> Thanks.

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

Please reply to the list and not to my personal email.
Those desiring my consulting or training services please feel free to email me.
---
"Nolite dare sanctum canibus neque mittatis margaritas vestras ante porcos ne forte conculcent eas pedibus suis."
Cum es damnatorum possederunt porcos iens ut salire off sanguinum cliff in abyssum?"

Bruce Weaver

Re: Geometric mean of a variable

Administrator

Maybe I'm missing something obvious, but why not use the MEAN function in the AGGREGATE instead of SUM and N? I.e.,

COMPUTE newvar=LG10(oldvar).
AGGREGATE OUTFILE * MODE=ADDVARIABLES / Mnewvar=MEAN(newvar).
COMPUTE GeoMean=10**(Mnewvar).

David Marso wrote

Close:
But
1. SPSS doesn't support LOG.
Use either LN or LG10.
LG10(numexpr)
Returns the base-10 logarithm of numexpr, which must be numeric and greater than 0.
LN(numexpr)
Returns the base-e logarithm of numexpr, which must be numeric and greater than 0.
2. You left out a final step (divide by 10).
---
So .. relevant code (untested)**.
--
COMPUTE newvar=LG10(oldvar).
AGGREGATE OUTFILE * MODE=ADDVARIABLES / Snewvar=SUM(newvar) / N=N(newvar).
COMPUTE GeoMean=10^(SNewvar/N).
---

Mike Palij wrote

Someone may be able to provide the relevant code but this is what you do:

(1) Create a new variable logvar001=log(var001).

(2) Create the sum of logvar001 (probably through aggregate and add to
active dataset).

(3) Compute the antilog of the sum of logvar001.

compute alogsum=10**sumlogvar001.

Here are examples of how to compute the Geometric mean by hand or whatever:
http://books.google.com/books?id=3C9LzilnF6YC&pg=PA83&dq=%22calculation+of+geometric+mean%22&hl=en&sa=X&ei=0_jEULSPN8i80QH28wE&ved=0CDMQ6AEwAQ#v=onepage&q=%22calculation%20of%20geometric%20mean%22&f=false

And here is an example of how to get an antilog in SPSS:
http://www-01.ibm.com/support/docview.wss?uid=swg21480979

I'm sure that if I'm error someone will correct me.

-Mike Palij
New York University
[hidden email]

----- Original Message -----
From: "drfg2008" <[hidden email]>
To: <[hidden email]>
Sent: Sunday, December 09, 2012 3:26 PM
Subject: Geometric mean of a variable

> SPSS 20
>
> How can I compute a geometric mean of a variable ?
> (The problem is, that the variable only consists of positive numbers LE 1
> and N of numbers is large -> 38.000)
>
> Example: What is the geometric mean of VAR001 ?
>
> input program.
> loop a =1 to 38*10**3 by 1.
> end case.
> end loop.
> end file.
> end input program.
> EXECUTE.
> COMPUTE VAR001 =RV.UNIFORM(0,1).
> EXECUTE.
>
> Thanks.

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

--
Bruce Weaver
bweaver@lakeheadu.ca
http://sites.google.com/a/lakeheadu.ca/bweaver/

"When all else fails, RTFM."

PLEASE NOTE THE FOLLOWING:
1. My Hotmail account is not monitored regularly. To send me an e-mail, please use the address shown above.
2. The SPSSX Discussion forum on Nabble is no longer linked to the SPSSX-L listserv administered by UGA (https://listserv.uga.edu/).

David Marso

Re: Geometric mean of a variable

Administrator

Because I just woke up from a nap ;-)
--

Bruce Weaver wrote

Maybe I'm missing something obvious, but why not use the MEAN function in the AGGREGATE instead of SUM and N? I.e.,

COMPUTE newvar=LG10(oldvar).
AGGREGATE OUTFILE * MODE=ADDVARIABLES / Mnewvar=MEAN(newvar).
COMPUTE GeoMean=10**(Mnewvar).

David Marso wrote

Close:
But
1. SPSS doesn't support LOG.
Use either LN or LG10.
LG10(numexpr)
Returns the base-10 logarithm of numexpr, which must be numeric and greater than 0.
LN(numexpr)
Returns the base-e logarithm of numexpr, which must be numeric and greater than 0.
2. You left out a final step (divide by 10).
---
So .. relevant code (untested)**.
--
COMPUTE newvar=LG10(oldvar).
AGGREGATE OUTFILE * MODE=ADDVARIABLES / Snewvar=SUM(newvar) / N=N(newvar).
COMPUTE GeoMean=10^(SNewvar/N).
---

Mike Palij wrote

Someone may be able to provide the relevant code but this is what you do:

(1) Create a new variable logvar001=log(var001).

(2) Create the sum of logvar001 (probably through aggregate and add to
active dataset).

(3) Compute the antilog of the sum of logvar001.

compute alogsum=10**sumlogvar001.

Here are examples of how to compute the Geometric mean by hand or whatever:
http://books.google.com/books?id=3C9LzilnF6YC&pg=PA83&dq=%22calculation+of+geometric+mean%22&hl=en&sa=X&ei=0_jEULSPN8i80QH28wE&ved=0CDMQ6AEwAQ#v=onepage&q=%22calculation%20of%20geometric%20mean%22&f=false

And here is an example of how to get an antilog in SPSS:
http://www-01.ibm.com/support/docview.wss?uid=swg21480979

I'm sure that if I'm error someone will correct me.

-Mike Palij
New York University
[hidden email]

----- Original Message -----
From: "drfg2008" <[hidden email]>
To: <[hidden email]>
Sent: Sunday, December 09, 2012 3:26 PM
Subject: Geometric mean of a variable

> SPSS 20
>
> How can I compute a geometric mean of a variable ?
> (The problem is, that the variable only consists of positive numbers LE 1
> and N of numbers is large -> 38.000)
>
> Example: What is the geometric mean of VAR001 ?
>
> input program.
> loop a =1 to 38*10**3 by 1.
> end case.
> end loop.
> end file.
> end input program.
> EXECUTE.
> COMPUTE VAR001 =RV.UNIFORM(0,1).
> EXECUTE.
>
> Thanks.

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

Mike

Re: Geometric mean of a variable

In reply to this post by Jon K Peck

On Sunday, December 09, 2012 7:07 PM, Jon K Peck wrote:
>The oldest version of SPSS I have installed is 15, and the geometric mean
>is
>available there. Since there is no update notice for it in the command
>history,
>it is probably present at least back to V13 when command histories were
>introduced.

I'll take your word for it. I have a PDF for algorithms for v12 and there
is no formula for the geometric mean given there.

>But why would you expect every descriptive procedure to offer all the same
>statistics, >especially ones that are little used?

It would be nice, especially for pedagogical purposes, to have a single
procedure
that would provide a comprehenive set of descriptive statistics, both common
and
rare. Frequencies was at one time the procedure for this but, in my opinion,
was
superceded by Examine/explore, especially in providing measures of central
tendency,
both standard and robust (which one could argue are also "little used"
unless, of
course, you use them). It would be nice if the geometric and harmonic means
were options in frequencies but they should definitively made options for
Examine/explore where one may want to examine several measures of central
tendency. If one covers the Generalized Mean formula, it would be nice to
have
the relevant means, see:
http://en.wikipedia.org/wiki/Generalized_mean

My own experience with geometric means goes back to graduate school where
the geometric mean was computed for psychophysical data being
analyzed/fitted
with Stevens' Power Law. The entry on the geometric mean on Wikipedia
provides a variety of examples/applications where it is relevant, so how
"little
used" it is might depend upon which discipline one is in; see:
http://en.wikipedia.org/wiki/Geometric_mean

-Mike Palij
New York University
[hidden email]

----- Original Message -----
From: Mike Palij <[hidden email]>
To: [hidden email],
Date: 12/09/2012 03:15 PM
Subject: Re: [SPSSX-L] Geometric mean of a variable
Sent by: "SPSSX(r) Discussion" <[hidden email]>

Now that is interesting. I double-checked the v20 manual and
see that both the means and summarize procedure provide the
geometric mean. I only have two questions:

(1) When did this become available?

and

(2) Why aren't these provided in frequencies, descriptives,
and examine/explore?

-Mike Palij
New York University
[hidden email]
----- Original Message -----
From: Art Kendall
To: [hidden email]
Sent: Sunday, December 09, 2012 3:57 PM
Subject: Re: Geometric mean of a variable

input program.
loop a =1 to 38*10**3 by 1.
COMPUTE VAR001 =RV.UNIFORM(0,1).
end case.
end loop.
end file.
end input program.
MEANS TABLES=VAR001
/CELLS MEAN COUNT STDDEV GEOMETRIC SUM.

VAR001
Mean N Std. Deviation Geometric Mean Sum
.50 38000 .289 .37 18996

Resources Processor Time 00:00:00.08
Elapsed Time 00:00:00.14

Art Kendall
Social Research Consultants
On 12/9/2012 3:26 PM, drfg2008 wrote:
input program.
loop a =1 to 38*10**3 by 1.
end case.
end loop.
end file.
end input program.
EXECUTE.
COMPUTE VAR001 =RV.UNIFORM(0,1).
EXECUTE.

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

Bruce Weaver

Re: Geometric mean of a variable

Administrator

I agree with Mike that it would be very nice/convenient to have all of the available descriptive stats as options for a single procedure. My preference would be to include them all in DESCRIPTIVES, first because the output is much tidier than for the other candidates (i.e., one row per variable with the selected descriptive stats in columns), and second, because students and other beginners who are trying to generate descriptive statistics find it confusing when they can't get what they want under Analyze > Descriptive Statistics > Descriptives, and instead have to look under Compare Means or Frequencies. (And given that the code is already written, this would seem a trivial change.)

There's my tuppence!

;-)

Mike Palij wrote

On Sunday, December 09, 2012 7:07 PM, Jon K Peck wrote:
>The oldest version of SPSS I have installed is 15, and the geometric mean
>is
>available there. Since there is no update notice for it in the command
>history,
>it is probably present at least back to V13 when command histories were
>introduced.

I'll take your word for it. I have a PDF for algorithms for v12 and there
is no formula for the geometric mean given there.

>But why would you expect every descriptive procedure to offer all the same
>statistics, >especially ones that are little used?

It would be nice, especially for pedagogical purposes, to have a single
procedure
that would provide a comprehenive set of descriptive statistics, both common
and
rare. Frequencies was at one time the procedure for this but, in my opinion,
was
superceded by Examine/explore, especially in providing measures of central
tendency,
both standard and robust (which one could argue are also "little used"
unless, of
course, you use them). It would be nice if the geometric and harmonic means
were options in frequencies but they should definitively made options for
Examine/explore where one may want to examine several measures of central
tendency. If one covers the Generalized Mean formula, it would be nice to
have
the relevant means, see:
http://en.wikipedia.org/wiki/Generalized_mean

My own experience with geometric means goes back to graduate school where
the geometric mean was computed for psychophysical data being
analyzed/fitted
with Stevens' Power Law. The entry on the geometric mean on Wikipedia
provides a variety of examples/applications where it is relevant, so how
"little
used" it is might depend upon which discipline one is in; see:
http://en.wikipedia.org/wiki/Geometric_mean

-Mike Palij
New York University
[hidden email]

----- Original Message -----
From: Mike Palij <[hidden email]>
To: [hidden email],
Date: 12/09/2012 03:15 PM
Subject: Re: [SPSSX-L] Geometric mean of a variable
Sent by: "SPSSX(r) Discussion" <[hidden email]>

Now that is interesting. I double-checked the v20 manual and
see that both the means and summarize procedure provide the
geometric mean. I only have two questions:

(1) When did this become available?

and

(2) Why aren't these provided in frequencies, descriptives,
and examine/explore?

-Mike Palij
New York University
[hidden email]
----- Original Message -----
From: Art Kendall
To: [hidden email]
Sent: Sunday, December 09, 2012 3:57 PM
Subject: Re: Geometric mean of a variable

input program.
loop a =1 to 38*10**3 by 1.
COMPUTE VAR001 =RV.UNIFORM(0,1).
end case.
end loop.
end file.
end input program.
MEANS TABLES=VAR001
/CELLS MEAN COUNT STDDEV GEOMETRIC SUM.

VAR001
Mean N Std. Deviation Geometric Mean Sum
.50 38000 .289 .37 18996

Resources Processor Time 00:00:00.08
Elapsed Time 00:00:00.14

Art Kendall
Social Research Consultants
On 12/9/2012 3:26 PM, drfg2008 wrote:
input program.
loop a =1 to 38*10**3 by 1.
end case.
end loop.
end file.
end input program.
EXECUTE.
COMPUTE VAR001 =RV.UNIFORM(0,1).
EXECUTE.

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

Jon K Peck

Re: Geometric mean of a variable

In reply to this post by Mike

I agree that a single descriptives procedure would be nice. IMO It should use the measurement level to sort out which statistics are appropriate for each variable. The CODEBOOK procedure is sensitive to the measurement level and gives you a unified view with sensible statistics, but it provides only very basic measures. Summarize has pretty much everything for continuous variables but nothing for categorical ones.

I sometimes teach a programmability class in which we construct a toy unified descriptives procedure as an extension command complete with a dialog box interface that produces ml-appropriate statistics by using both FREQUENCIES and SUMMARIZE or EXAMINE. It would be nice to have something like that built in, or even a full fledged extension command for this.

As for adding generalized and harmonic means to EXAMINE, send a request to [hidden email]. No promises, especially since the EXAMINE code is pretty complex and difficult to work with, but it could get on the radar for future developments.

Jon Peck (no "h") aka Kim
Senior Software Engineer, IBM
[hidden email]
new phone: 720-342-5621

From: Mike Palij <[hidden email]>
To: [hidden email],
Date: 12/09/2012 08:08 PM
Subject: Re: [SPSSX-L] Geometric mean of a variable
Sent by: "SPSSX(r) Discussion" <[hidden email]>

On Sunday, December 09, 2012 7:07 PM, Jon K Peck wrote: >The oldest version of SPSS I have installed is 15, and the geometric mean >is >available there. Since there is no update notice for it in the command >history, >it is probably present at least back to V13 when command histories were >introduced. I'll take your word for it. I have a PDF for algorithms for v12 and there is no formula for the geometric mean given there. >But why would you expect every descriptive procedure to offer all the same >statistics, >especially ones that are little used? It would be nice, especially for pedagogical purposes, to have a single procedure that would provide a comprehenive set of descriptive statistics, both common and rare. Frequencies was at one time the procedure for this but, in my opinion, was superceded by Examine/explore, especially in providing measures of central tendency, both standard and robust (which one could argue are also "little used" unless, of course, you use them). It would be nice if the geometric and harmonic means were options in frequencies but they should definitively made options for Examine/explore where one may want to examine several measures of central tendency. If one covers the Generalized Mean formula, it would be nice to have the relevant means, see:http://en.wikipedia.org/wiki/Generalized_meanMy own experience with geometric means goes back to graduate school where the geometric mean was computed for psychophysical data being analyzed/fitted with Stevens' Power Law. The entry on the geometric mean on Wikipedia provides a variety of examples/applications where it is relevant, so how "little used" it is might depend upon which discipline one is in; see:http://en.wikipedia.org/wiki/Geometric_mean-Mike Palij New York University [hidden email] ----- Original Message ----- From: Mike Palij <[hidden email]> To: [hidden email], Date: 12/09/2012 03:15 PM Subject: Re: [SPSSX-L] Geometric mean of a variable Sent by: "SPSSX(r) Discussion" <[hidden email]> Now that is interesting. I double-checked the v20 manual and see that both the means and summarize procedure provide the geometric mean. I only have two questions: (1) When did this become available? and (2) Why aren't these provided in frequencies, descriptives, and examine/explore? -Mike Palij New York University [hidden email] ----- Original Message ----- From: Art Kendall To: [hidden email] Sent: Sunday, December 09, 2012 3:57 PM Subject: Re: Geometric mean of a variable input program. loop a =1 to 38*10**3 by 1. COMPUTE VAR001 =RV.UNIFORM(0,1). end case. end loop. end file. end input program. MEANS TABLES=VAR001 /CELLS MEAN COUNT STDDEV GEOMETRIC SUM. VAR001 Mean N Std. Deviation Geometric Mean Sum .50 38000 .289 .37 18996 Resources Processor Time 00:00:00.08 Elapsed Time 00:00:00.14 Art Kendall Social Research Consultants On 12/9/2012 3:26 PM, drfg2008 wrote: input program. loop a =1 to 38*10**3 by 1. end case. end loop. end file. end input program. EXECUTE. COMPUTE VAR001 =RV.UNIFORM(0,1). EXECUTE. ===================== 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

Bruce Weaver

Re: Geometric mean of a variable

Administrator

In reply to this post by Ryan

Instead of using UNIANOVA, you could do this:

* Generalized Linear Models.
GENLIN natural_log_VAR001
/MODEL INTERCEPT=YES
DISTRIBUTION=NORMAL LINK=IDENTITY
/MISSING CLASSMISSING=EXCLUDE
/PRINT SOLUTION (EXPONENTIATED).

Exp(B) = the geometric mean for VAR001.

I tried other variations with Y = VAR001 and a Log link-function (for example); but I couldn't get it to work out properly. Perhaps I wasn't persistent (or clever) enough. ;-)

HTH.

R B wrote

By fitting an intercept-only general linear model on the natural log of the
variable and then exponentiating the estimated intercept and confidence
limits will provide both the geometric mean and the 95% confidence
interval, respectively:

input program.
loop a =1 to 38*10**3 by 1.
end case.
end loop.
end file.
end input program.
EXECUTE.
COMPUTE VAR001 =RV.UNIFORM(0,1).
EXECUTE.

compute natural_log_VAR001=ln(VAR001).
execute.

UNIANOVA natural_log_VAR001
/METHOD=SSTYPE(3)
/INTERCEPT=INCLUDE
/PRINT=PARAMETER
/CRITERIA=ALPHA(.05).

compute geometric_mean = exp(-.9951088706552569).
compute geometric_mean_LL = exp(-1.005089860177377).
compute geometric_mean_UL = exp(-.9851278811331369).
execute.

Ryan

On Sun, Dec 9, 2012 at 3:26 PM, drfg2008 <[hidden email]> wrote:
> SPSS 20
>
> How can I compute a geometric mean of a variable ?
> (The problem is, that the variable only consists of positive numbers LE 1
> and N of numbers is large -> 38.000)
>
> Example: What is the geometric mean of VAR001 ?
>
> input program.
> loop a =1 to 38*10**3 by 1.
> end case.
> end loop.
> end file.
> end input program.
> EXECUTE.
> COMPUTE VAR001 =RV.UNIFORM(0,1).
> EXECUTE.
>
> Thanks.
>
>
>
> -----
> Dr. Frank Gaeth
> FU-Berlin
>
> --
> View this message in context:
http://spssx-discussion.1045642.n5.nabble.com/Geometric-mean-of-a-variable-tp5716795.html
> Sent from the SPSSX Discussion mailing list archive at Nabble.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

Ryan

Re: Geometric mean of a variable

It didn't work because a model with a specified log normal distribution with an identity link is not equivalent to a model with specified normal distribution and a log link.

Ryan

On Mon, Dec 10, 2012 at 11:30 AM, Bruce Weaver <[hidden email]> wrote:

Instead of using UNIANOVA, you could do this:

* Generalized Linear Models.
GENLIN natural_log_VAR001
/MODEL INTERCEPT=YES
DISTRIBUTION=NORMAL LINK=IDENTITY
/MISSING CLASSMISSING=EXCLUDE
/PRINT SOLUTION (EXPONENTIATED).

Exp(B) = the geometric mean for VAR001.

I tried other variations with Y = VAR001 and a Log link-function (for
example); but I couldn't get it to work out properly. Perhaps I wasn't
persistent (or clever) enough. ;-)

HTH.

R B wrote

> By fitting an intercept-only general linear model on the natural log of
> the
> variable and then exponentiating the estimated intercept and confidence
> limits will provide both the geometric mean and the 95% confidence
> interval, respectively:
>
> input program.
> loop a =1 to 38*10**3 by 1.
> end case.
> end loop.
> end file.
> end input program.
> EXECUTE.
> COMPUTE VAR001 =RV.UNIFORM(0,1).
> EXECUTE.
>
> compute natural_log_VAR001=ln(VAR001).
> execute.
>
> UNIANOVA natural_log_VAR001
> /METHOD=SSTYPE(3)
> /INTERCEPT=INCLUDE
> /PRINT=PARAMETER
> /CRITERIA=ALPHA(.05).
>
> compute geometric_mean = exp(-.9951088706552569).
> compute geometric_mean_LL = exp(-1.005089860177377).
> compute geometric_mean_UL = exp(-.9851278811331369).
> execute.
>
> Ryan
>
>
> On Sun, Dec 9, 2012 at 3:26 PM, drfg2008 <

> kontakt@

> > wrote:
>> SPSS 20
>>
>> How can I compute a geometric mean of a variable ?
>> (The problem is, that the variable only consists of positive numbers LE 1
>> and N of numbers is large -> 38.000)
>>
>> Example: What is the geometric mean of VAR001 ?
>>
>> input program.
>> loop a =1 to 38*10**3 by 1.
>> end case.
>> end loop.
>> end file.
>> end input program.
>> EXECUTE.
>> COMPUTE VAR001 =RV.UNIFORM(0,1).
>> EXECUTE.
>>
>> Thanks.
>>
>>
>>
>> -----
>> Dr. Frank Gaeth
>> FU-Berlin
>>
>> --
>> View this message in context:
> http://spssx-discussion.1045642.n5.nabble.com/Geometric-mean-of-a-variable-tp5716795.html
>> Sent from the SPSSX Discussion mailing list archive at Nabble.com.
>>
>> =====================
>> To manage your subscription to SPSSX-L, send a message to
>>

> LISTSERV@.UGA

> (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

-----
--
Bruce Weaver
[hidden email]
http://sites.google.com/a/lakeheadu.ca/bweaver/

"When all else fails, RTFM."

NOTE: My Hotmail account is not monitored regularly.
To send me an e-mail, please use the address shown above.

--
View this message in context: http://spssx-discussion.1045642.n5.nabble.com/Geometric-mean-of-a-variable-tp5716795p5716825.html

Sent from the SPSSX Discussion mailing list archive at Nabble.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

Bruce Weaver

Re: Geometric mean of a variable

Administrator

Just to be clear, when Ryan says "it didn't work", he is referring to the "other variations" I mentioned. The GENLIN command shown in my post *did* work.

HTH.

R B wrote

It didn't work because a model with a specified log normal distribution
with an identity link is not equivalent to a model with specified normal
distribution and a log link.
Ryan
On Mon, Dec 10, 2012 at 11:30 AM, Bruce Weaver <[hidden email]>wrote:

> Instead of using UNIANOVA, you could do this:
>
> * Generalized Linear Models.
> GENLIN natural_log_VAR001
> /MODEL INTERCEPT=YES
> DISTRIBUTION=NORMAL LINK=IDENTITY
> /MISSING CLASSMISSING=EXCLUDE
> /PRINT SOLUTION (EXPONENTIATED).
>
> Exp(B) = the geometric mean for VAR001.
>
> I tried other variations with Y = VAR001 and a Log link-function (for
> example); but I couldn't get it to work out properly. Perhaps I wasn't
> persistent (or clever) enough. ;-)
>
> HTH.
>
>
>
> R B wrote
> > By fitting an intercept-only general linear model on the natural log of
> > the
> > variable and then exponentiating the estimated intercept and confidence
> > limits will provide both the geometric mean and the 95% confidence
> > interval, respectively:
> >
> > input program.
> > loop a =1 to 38*10**3 by 1.
> > end case.
> > end loop.
> > end file.
> > end input program.
> > EXECUTE.
> > COMPUTE VAR001 =RV.UNIFORM(0,1).
> > EXECUTE.
> >
> > compute natural_log_VAR001=ln(VAR001).
> > execute.
> >
> > UNIANOVA natural_log_VAR001
> > /METHOD=SSTYPE(3)
> > /INTERCEPT=INCLUDE
> > /PRINT=PARAMETER
> > /CRITERIA=ALPHA(.05).
> >
> > compute geometric_mean = exp(-.9951088706552569).
> > compute geometric_mean_LL = exp(-1.005089860177377).
> > compute geometric_mean_UL = exp(-.9851278811331369).
> > execute.
> >
> > Ryan
> >
> >
> > On Sun, Dec 9, 2012 at 3:26 PM, drfg2008 <
>
> > kontakt@
>
> > > wrote:
> >> SPSS 20
> >>
> >> How can I compute a geometric mean of a variable ?
> >> (The problem is, that the variable only consists of positive numbers LE
> 1
> >> and N of numbers is large -> 38.000)
> >>
> >> Example: What is the geometric mean of VAR001 ?
> >>
> >> input program.
> >> loop a =1 to 38*10**3 by 1.
> >> end case.
> >> end loop.
> >> end file.
> >> end input program.
> >> EXECUTE.
> >> COMPUTE VAR001 =RV.UNIFORM(0,1).
> >> EXECUTE.
> >>
> >> Thanks.
> >>
> >>
> >>
> >> -----
> >> Dr. Frank Gaeth
> >> FU-Berlin
> >>
> >> --
> >> View this message in context:
> >
> http://spssx-discussion.1045642.n5.nabble.com/Geometric-mean-of-a-variable-tp5716795.html
> >> Sent from the SPSSX Discussion mailing list archive at Nabble.com.
> >>
> >> =====================
> >> To manage your subscription to SPSSX-L, send a message to
> >>
>
> > LISTSERV@.UGA
>
> > (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
>
>
>
>
>
> -----
> --
> Bruce Weaver
> [hidden email]
> http://sites.google.com/a/lakeheadu.ca/bweaver/
>
> "When all else fails, RTFM."
>
> NOTE: My Hotmail account is not monitored regularly.
> To send me an e-mail, please use the address shown above.
>
> --
> View this message in context:
> http://spssx-discussion.1045642.n5.nabble.com/Geometric-mean-of-a-variable-tp5716795p5716825.html
> Sent from the SPSSX Discussion mailing list archive at Nabble.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
>

Ryan

Re: Geometric mean of a variable

Right. Apologies if I wasn't clear. -Ryan

On Dec 10, 2012, at 5:59 PM, Bruce Weaver <[hidden email]> wrote:

> Just to be clear, when Ryan says "it didn't work", he is referring to the
> "other variations" I mentioned. The GENLIN command shown in my post *did*
> work.
>
> HTH.
>
>
>
> R B wrote
>> It didn't work because a model with a specified log normal distribution
>> with an identity link is not equivalent to a model with specified normal
>> distribution and a log link.
>> Ryan
>> On Mon, Dec 10, 2012 at 11:30 AM, Bruce Weaver <
>
>> bruce.weaver@
>
>> >wrote:
>>
>>> Instead of using UNIANOVA, you could do this:
>>>
>>> * Generalized Linear Models.
>>> GENLIN natural_log_VAR001
>>> /MODEL INTERCEPT=YES
>>> DISTRIBUTION=NORMAL LINK=IDENTITY
>>> /MISSING CLASSMISSING=EXCLUDE
>>> /PRINT SOLUTION (EXPONENTIATED).
>>>
>>> Exp(B) = the geometric mean for VAR001.
>>>
>>> I tried other variations with Y = VAR001 and a Log link-function (for
>>> example); but I couldn't get it to work out properly. Perhaps I wasn't
>>> persistent (or clever) enough. ;-)
>>>
>>> HTH.
>>>
>>>
>>>
>>> R B wrote
>>>> By fitting an intercept-only general linear model on the natural log of
>>>> the
>>>> variable and then exponentiating the estimated intercept and confidence
>>>> limits will provide both the geometric mean and the 95% confidence
>>>> interval, respectively:
>>>>
>>>> input program.
>>>> loop a =1 to 38*10**3 by 1.
>>>> end case.
>>>> end loop.
>>>> end file.
>>>> end input program.
>>>> EXECUTE.
>>>> COMPUTE VAR001 =RV.UNIFORM(0,1).
>>>> EXECUTE.
>>>>
>>>> compute natural_log_VAR001=ln(VAR001).
>>>> execute.
>>>>
>>>> UNIANOVA natural_log_VAR001
>>>> /METHOD=SSTYPE(3)
>>>> /INTERCEPT=INCLUDE
>>>> /PRINT=PARAMETER
>>>> /CRITERIA=ALPHA(.05).
>>>>
>>>> compute geometric_mean = exp(-.9951088706552569).
>>>> compute geometric_mean_LL = exp(-1.005089860177377).
>>>> compute geometric_mean_UL = exp(-.9851278811331369).
>>>> execute.
>>>>
>>>> Ryan
>>>>
>>>>
>>>> On Sun, Dec 9, 2012 at 3:26 PM, drfg2008 <
>>>
>>>> kontakt@
>>>
>>>>> wrote:
>>>>> SPSS 20
>>>>>
>>>>> How can I compute a geometric mean of a variable ?
>>>>> (The problem is, that the variable only consists of positive numbers
>>> LE
>>> 1
>>>>> and N of numbers is large -> 38.000)
>>>>>
>>>>> Example: What is the geometric mean of VAR001 ?
>>>>>
>>>>> input program.
>>>>> loop a =1 to 38*10**3 by 1.
>>>>> end case.
>>>>> end loop.
>>>>> end file.
>>>>> end input program.
>>>>> EXECUTE.
>>>>> COMPUTE VAR001 =RV.UNIFORM(0,1).
>>>>> EXECUTE.
>>>>>
>>>>> Thanks.
>>>>>
>>>>>
>>>>>
>>>>> -----
>>>>> Dr. Frank Gaeth
>>>>> FU-Berlin
>>>>>
>>>>> --
>>>>> View this message in context:
>>> http://spssx-discussion.1045642.n5.nabble.com/Geometric-mean-of-a-variable-tp5716795.html
>>>>> Sent from the SPSSX Discussion mailing list archive at Nabble.com.
>>>>>
>>>>> =====================
>>>>> To manage your subscription to SPSSX-L, send a message to
>>>
>>>> LISTSERV@.UGA
>>>
>>>> (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
>>>
>>>
>>>
>>>
>>>
>>> -----
>>> --
>>> Bruce Weaver
>
>> bweaver@
>
>>> http://sites.google.com/a/lakeheadu.ca/bweaver/
>>>
>>> "When all else fails, RTFM."
>>>
>>> NOTE: My Hotmail account is not monitored regularly.
>>> To send me an e-mail, please use the address shown above.
>>>
>>> --
>>> View this message in context:
>>> http://spssx-discussion.1045642.n5.nabble.com/Geometric-mean-of-a-variable-tp5716795p5716825.html
>>> Sent from the SPSSX Discussion mailing list archive at Nabble.com.
>>>
>>> =====================
>>> To manage your subscription to SPSSX-L, send a message to
>
>> LISTSERV@.UGA
>
>> (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
>
>
>
>
>
> -----
> --
> Bruce Weaver
> [hidden email]
> http://sites.google.com/a/lakeheadu.ca/bweaver/
>
> "When all else fails, RTFM."
>
> NOTE: My Hotmail account is not monitored regularly.
> To send me an e-mail, please use the address shown above.
>
> --
> View this message in context: http://spssx-discussion.1045642.n5.nabble.com/Geometric-mean-of-a-variable-tp5716795p5716849.html
> Sent from the SPSSX Discussion mailing list archive at Nabble.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

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