Pooled data and Logistic Regression

[jaya609]

CONTENTS DELETED

The author has deleted this message.

[Bruce Weaver]

Dare I ask why you want to dichotomize?  More often than not, it's a bad idea:  You throw away information and lose power.  There are several good articles on the topic, including this very readable one by Dave Streiner.

  https://ww1.cpa-apc.org/publications/archives/cjp/2002/april/streiner.PDF

HTH.

jaya609 wrote

Dear Listserv,

I used Multiple Imputation in SPSS 17 MVA.
Now, I want to dichotomize my DVs based on their median value.
However, in frequencies output, I did not see median for POOLED dataset.
How can I get it?

The next step is Logistic Regression.
Some tables of Logistic Regression output.... I did not get the result of pooled dataset.
Those tables are...
Omnibus tests of model coefficients, Model summary, and Classification table.
Is there any way to get the results of pooled dataset on these tables?
If it's not.... how can I report my results?

JHK

[Art Kendall]

Once I saw two actual occasions when dichotomizing variables was useful.

Back when logistic regression was new a presenter wanted to describe what logistic regression was. She did not have a data set with a dichotomous DV so she used an existing data set. [This was late 70's, before the web, so it was not easy to find an appropriate data set quickly.]  Unfortunately, the audience did not grasp immediately that she was just doing a demo of LR itself, and lambasted her for "committing a nefarious median split". That ate up a lot of the time for her presentation.

On another occasion a presenter wanted to demonstrate the futility of the median split. 

Have you experienced any occasion where it made sense to dichotomize?

Art

On 4/4/2011 9:05 AM, Bruce Weaver wrote:

Dare I ask why you want to dichotomize?  More often than not, it's a bad
idea:  You throw away information and lose power.  There are several good
articles on the topic, including this very readable one by Dave Streiner.

  https://ww1.cpa-apc.org/publications/archives/cjp/2002/april/streiner.PDF

HTH.



jaya609 wrote:

Dear Listserv,

I used Multiple Imputation in SPSS 17 MVA.
Now, I want to dichotomize my DVs based on their median value.
However, in frequencies output, I did not see median for POOLED dataset.
How can I get it?

The next step is Logistic Regression.
Some tables of Logistic Regression output.... I did not get the result of
pooled dataset.
Those tables are...
Omnibus tests of model coefficients, Model summary, and Classification
table.
Is there any way to get the results of pooled dataset on these tables?
If it's not.... how can I report my results?

JHK

-----
--
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/Pooled-data-and-Logistic-Regression-tp4279890p4281707.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

[Art Kendall]

Oops.  Did not intend that to go to the whole list.

Art

On 4/4/2011 9:28 AM, Art Kendall wrote:

Once I saw two actual occasions when dichotomizing variables was useful.

Back when logistic regression was new a presenter wanted to describe what logistic regression was. She did not have a data set with a dichotomous DV so she used an existing data set. [This was late 70's, before the web, so it was not easy to find an appropriate data set quickly.]  Unfortunately, the audience did not grasp immediately that she was just doing a demo of LR itself, and lambasted her for "committing a nefarious median split". That ate up a lot of the time for her presentation.

On another occasion a presenter wanted to demonstrate the futility of the median split. 

Have you experienced any occasion where it made sense to dichotomize?

Art

On 4/4/2011 9:05 AM, Bruce Weaver wrote:

Dare I ask why you want to dichotomize?  More often than not, it's a bad
idea:  You throw away information and lose power.  There are several good
articles on the topic, including this very readable one by Dave Streiner.

  https://ww1.cpa-apc.org/publications/archives/cjp/2002/april/streiner.PDF

HTH.



jaya609 wrote:

Dear Listserv,

I used Multiple Imputation in SPSS 17 MVA.
Now, I want to dichotomize my DVs based on their median value.
However, in frequencies output, I did not see median for POOLED dataset.
How can I get it?

The next step is Logistic Regression.
Some tables of Logistic Regression output.... I did not get the result of
pooled dataset.
Those tables are...
Omnibus tests of model coefficients, Model summary, and Classification
table.
Is there any way to get the results of pooled dataset on these tables?
If it's not.... how can I report my results?

JHK

-----
--
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/Pooled-data-and-Logistic-Regression-tp4279890p4281707.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

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

[jaya609]

CONTENTS DELETED

The author has deleted this message.

[Swank, Paul R]

There are better alternatives to dichotomizing for skewed data. Generalized linear models might be an option for positively skewed data and negatively skewed can usually be converted to a positive skew easily enough. Then a Poisson or negative binomial distribution with a log link function should handle it.

Dr. Paul R. Swank,
Professor and Director of Research
Children's Learning Institute
University of Texas Health Science Center-Houston


-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of jaya609
Sent: Monday, April 04, 2011 11:43 AM
To: [hidden email]
Subject: Re: Pooled data and Logistic Regression

I know if I dichotomize and make Dvs as binary, I loose information and
power.
However, the DVs are quite skewed and there's no way around it.
In order to ask my theoretical question, running LR after dichotomize Dvs.

Is there anybody who can answer my original question regarding Pooled data
and output in SPSS LR?

--
View this message in context: http://spssx-discussion.1045642.n5.nabble.com/Pooled-data-and-Logistic-Regression-tp4279890p4282210.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

[Art Kendall]

Are the residuals skewed or is it just the raw data that is skewed?
Did you check that the extreme values are not data entry errors?

Can you think of a transform that would that would make substantive sense in your problem?

Art Kendall
Social Research Consultants

On 4/4/2011 12:43 PM, jaya609 wrote:

I know if I dichotomize and make Dvs as binary, I loose information and
power.
However, the DVs are quite skewed and there's no way around it.
In order to ask my theoretical question, running LR after dichotomize Dvs.

Is there anybody who can answer my original question regarding Pooled data
and output in SPSS LR?

--
View this message in context: http://spssx-discussion.1045642.n5.nabble.com/Pooled-data-and-Logistic-Regression-tp4279890p4282210.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