how to handle absolute zero values?

[lcl23]

There are 2 variables (ratio scale) in my data which have more than 15% of absolute zero values, i.e. 0= no loan & 0= no asset. While entering the data into SPSS, how should I deal with these absolute zero values? Should I define all zeros as "missing values", which I think it might change the meaning of these zeros? Or should I not doing anything and leave the data as it is?

I am going to analyse the data using logistic regression method. Thanks.

[Bruce Weaver]

lcl23 wrote

There are 2 variables (ratio scale) in my data which have more than 15% of absolute zero values, i.e. 0= no loan & 0= no asset. While entering the data into SPSS, how should I deal with these absolute zero values? Should I define all zeros as "missing values", which I think it might change the meaning of these zeros? Or should I not doing anything and leave the data as it is?

I am going to analyse the data using logistic regression method. Thanks.

I don't understand where the problem is.  Unlike discriminant function analysis, logistic regression makes no assumptions about how explanatory variables are distributed.  

[lcl23]

I am not concerning about the distribution of variables whether they are normal or not. I am still in the process of cleaning the data before going into logistic regression. Would like to look for the correct treatment of these zero values. 

--- On Sun, 13/3/11, Bruce Weaver [via SPSSX Discussion] <[hidden email]> wrote:

From: Bruce Weaver [via SPSSX Discussion] <[hidden email]>
Subject: Re: how to handle absolute zero values?
To: "lcl23" <[hidden email]>
Date: Sunday, 13 March, 2011, 9:25 AM

lcl23 wrote:

There are 2 variables (ratio scale) in my data which have more than 15% of absolute zero values, i.e. 0= no loan & 0= no asset. While entering the data into SPSS, how should I deal with these absolute zero values? Should I define all zeros as "missing values", which I think it might change the meaning of these zeros? Or should I not doing anything and leave the data as it is?

I am going to analyse the data using logistic regression method. Thanks.

I don't understand where the problem is.  Unlike discriminant function analysis, logistic regression makes no assumptions about how explanatory variables are distributed.  

---

If you reply to this email, your message will be added to the discussion below:

http://spssx-discussion.1045642.n5.nabble.com/how-to-handle-absolute-zero-values-tp3451505p3545487.html

To unsubscribe from how to handle absolute zero values?, click here.

[Bruce Weaver]

Treating 0 as missing for those variables will result in folks with zeroes being excluded from your model.  What population are you trying to make an inference about?  Does it include or exclude people who have no loan, or no assets?

lcl23 wrote

I am not concerning about the distribution of variables whether they are normal or not. I am still in the process of cleaning the data before going into logistic regression. Would like to look for the correct treatment of these zero values. 

--- On Sun, 13/3/11, Bruce Weaver [via SPSSX Discussion] <[e-mail]> wrote:

From: Bruce Weaver [via SPSSX Discussion] <[e-mail]>
Subject: Re: how to handle absolute zero values?
To: "lcl23" <[e-mail]>
Date: Sunday, 13 March, 2011, 9:25 AM



       
lcl23 wrote:
There are 2 variables (ratio scale) in my data which have more than 15% of absolute zero values, i.e. 0= no loan & 0= no asset. While entering the data into SPSS, how should I deal with these absolute zero values? Should I define all zeros as "missing values", which I think it might change the meaning of these zeros? Or should I not doing anything and leave the data as it is?


I am going to analyse the data using logistic regression method. Thanks.



I don't understand where the problem is.  Unlike discriminant function analysis, logistic regression makes no assumptions about how explanatory variables are distributed.  



        --

Bruce Weaver

[e-mail]

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.


       

       

       
       
                If you reply to this email, your message will be added to the discussion below:
                http://spssx-discussion.1045642.n5.nabble.com/how-to-handle-absolute-zero-values-tp3451505p3545487.html
       
       
               
                To unsubscribe from how to handle absolute zero values?, click here.

[Ryan]

Do you have two binary variables or two continuous variables? If the
latter, what units are they in? Approximately what do the shapes of
the distributions look like? Do they appear zero-inflated? What are
the research questions you're hoping to answer? How were the data
collected? Do you have a natural hierarchy in your data (e.g., persons
nested in families)?

Please provide more details.

Ryan

On Sat, Mar 12, 2011 at 8:25 PM, Bruce Weaver <[hidden email]> wrote:

> lcl23 wrote:
>>
>> There are 2 variables (ratio scale) in my data which have more than 15% of
>> absolute zero values, i.e. 0= no loan & 0= no asset. While entering the
>> data into SPSS, how should I deal with these absolute zero values? Should
>> I define all zeros as "missing values", which I think it might change the
>> meaning of these zeros? Or should I not doing anything and leave the data
>> as it is?
>>
>> I am going to analyse the data using logistic regression method. Thanks.
>>
>
>
> I don't understand where the problem is.  Unlike discriminant function
> analysis, logistic regression makes no assumptions about how explanatory
> variables are distributed.
>
>
>
> -----
> --
> 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/how-to-handle-absolute-zero-values-tp3451505p3545487.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

[lcl23]

Dear Ryan,

1) Type of data: continuous 

2) Unit of measurement: (1) leverage: percentage; (2) directorship: integer

3) Skewness: (1) leverage: 1.94; (2) directorship: 1.42

4) Research question: (1) is the level of leverage affects company's decision to perform..... (2) is the number of directorship affects company's decision to perform.....

5) Zero values : (1) leverage: 19% of total sample ; (2) directorship: 15%

6) Data collection: secondary data from company annual reports

--- On Sun, 13/3/11, R B [via SPSSX Discussion] <[hidden email]> wrote:

From: R B [via SPSSX Discussion] <[hidden email]>
Subject: Re: how to handle absolute zero values?
To: "lcl23" <[hidden email]>
Date: Sunday, 13 March, 2011, 10:22 PM

Do you have two binary variables or two continuous variables? If the
latter, what units are they in? Approximately what do the shapes of
the distributions look like? Do they appear zero-inflated? What are
the research questions you're hoping to answer? How were the data
collected? Do you have a natural hierarchy in your data (e.g., persons
nested in families)?

Please provide more details.

Ryan

On Sat, Mar 12, 2011 at 8:25 PM, Bruce Weaver <[hidden email]> wrote:

> lcl23 wrote:
>>
>> There are 2 variables (ratio scale) in my data which have more than 15% of
>> absolute zero values, i.e. 0= no loan & 0= no asset. While entering the
>> data into SPSS, how should I deal with these absolute zero values? Should
>> I define all zeros as "missing values", which I think it might change the
>> meaning of these zeros? Or should I not doing anything and leave the data
>> as it is?
>>
>> I am going to analyse the data using logistic regression method. Thanks.
>>
>
>
> I don't understand where the problem is.  Unlike discriminant function
> analysis, logistic regression makes no assumptions about how explanatory
> variables are distributed.
>
>
>
> -----
> --
> 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/how-to-handle-absolute-zero-values-tp3451505p3545487.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

---

If you reply to this email, your message will be added to the discussion below:

http://spssx-discussion.1045642.n5.nabble.com/how-to-handle-absolute-zero-values-tp3451505p3555466.html

To unsubscribe from how to handle absolute zero values?, click here.

[Rich Ulrich]

When "zero"  is nothing like being an equal-interval
extension of the other numbers observed, then the
proper way to model either score is to use two variables:
the new variable will be an indicator variable, 0/1  for
No/Yes, which indicates that "directors" (say)  exist.
With both variables going into the model, it does not actually
matter what value is used for the sort-of-missing score.

In some cases it is easier to note the effects if the
Missing is assigned the mean of the rest.

The same solution can work when a zero=no-event  exists
for data that otherwise seem appropriate for using the
log transform or a reciprocal:  Use the transform, and
set the missing to the mean of the rest.

--
Rich Ulrich

________________________________

> Date: Sun, 13 Mar 2011 09:03:41 -0700
> From: [hidden email]
> Subject: Re: how to handle absolute zero values?
> To: [hidden email]
>
> Dear Ryan,
>
> 1) Type of data: continuous
> 2) Unit of measurement: (1) leverage: percentage; (2) directorship: integer
> 3) Skewness: (1) leverage: 1.94; (2) directorship: 1.42
> 4) Research question: (1) is the level of leverage affects company's
> decision to perform..... (2) is the number of directorship affects
> company's decision to perform.....
> 5) Zero values : (1) leverage: 19% of total sample ; (2) directorship: 15%
> 6) Data collection: secondary data from company annual reports
>

[snip, previous]


=====================
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]

Rich is describing Cohen & Cohen's "indicator" method for dealing with missing data.  It is no longer viewed favorably for situations where the data are truly missing (e.g., http://people.oregonstate.edu/~acock/growth-curves/working%20with%20missing%20values.pdf).  But that is not the case here--the zeroes are legitimate values.  I don't recall ever reading anything that made the distinction between truly missing and the kind of gap present in this case.  But the indicator method might be all right here.

Bruce

Rich Ulrich wrote

When "zero"  is nothing like being an equal-interval
extension of the other numbers observed, then the
proper way to model either score is to use two variables:
the new variable will be an indicator variable, 0/1  for
No/Yes, which indicates that "directors" (say)  exist.
With both variables going into the model, it does not actually
matter what value is used for the sort-of-missing score.

In some cases it is easier to note the effects if the
Missing is assigned the mean of the rest.

The same solution can work when a zero=no-event  exists
for data that otherwise seem appropriate for using the
log transform or a reciprocal:  Use the transform, and
set the missing to the mean of the rest.

--
Rich Ulrich

________________________________
> Date: Sun, 13 Mar 2011 09:03:41 -0700
> From: [e-mail]
> Subject: Re: how to handle absolute zero values?
> To: [e-mail]
>
> Dear Ryan,
>
> 1) Type of data: continuous
> 2) Unit of measurement: (1) leverage: percentage; (2) directorship: integer
> 3) Skewness: (1) leverage: 1.94; (2) directorship: 1.42
> 4) Research question: (1) is the level of leverage affects company's
> decision to perform..... (2) is the number of directorship affects
> company's decision to perform.....
> 5) Zero values : (1) leverage: 19% of total sample ; (2) directorship: 15%
> 6) Data collection: secondary data from company annual reports
>
[snip, previous]


=====================
To manage your subscription to SPSSX-L, send a message to
[e-mail] (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

[Rich Ulrich]

> Date: Mon, 14 Mar 2011 08:52:24 -0700
> From: [hidden email]
> Subject: Re: how to handle absolute zero values?
> To: [hidden email]
>
> Rich is describing Cohen & Cohen's "indicator" method for dealing with
> missing data. It is no longer viewed favorably for situations where the
> data are truly missing (e.g.,
> http://people.oregonstate.edu/~acock/growth-curves/working%20with%20missing%20values.pdf).

That article is (properly) harsh on the strategy of
simply substituting the mean for Missing.  It reviews
that strategy, and strategies for deletion and Imputation.

It does *not*  review the "indicator" method, though
it mentions it in passing, and mentions difficulties
when several indicators are highly correlated.  What the
article says about mean-substitution yielding reduced
standard deviations (owing to the inclusion of a set of
zero-variance mean-scores) -- should be kept in mind.

The article is a pretty good pragmatic overview.  It was
not sensitive to the important topic of how *many* Missings
have to be accounted for.


> But that is not the case here--the zeroes are legitimate values. I don't
> recall ever reading anything that made the distinction between truly missing
> and the kind of gap present in this case. But the indicator method might be
> all right here.
>
> Bruce

I say that it is a logical problem.  If the zero is not a
natural extension of the scale, and you care about the
effects of this variable, then you either use an Indicator

to allow for separate means, or you separate the sample
into two parts (zeroes, vs. others) to allow for entirely
separate regressions.

>
>
> Rich Ulrich wrote:
> >
> > When "zero" is nothing like being an equal-interval
> > extension of the other numbers observed, then the
> > proper way to model either score is to use two variables:
> > the new variable will be an indicator variable, 0/1 for
> > No/Yes, which indicates that "directors" (say) exist.
> > With both variables going into the model, it does not actually
> > matter what value is used for the sort-of-missing score.
> >
> > In some cases it is easier to note the effects if the
> > Missing is assigned the mean of the rest.
> >
> > The same solution can work when a zero=no-event exists
> > for data that otherwise seem appropriate for using the
> > log transform or a reciprocal: Use the transform, and
> > set the missing to the mean of the rest.
> >
> > --
> > Rich Ulrich

[snip, previous]



=====================
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]

What led you to consider converting zeroes to missing values? What's the issue?

 

Ryan

On Sun, Mar 13, 2011 at 12:03 PM, lcl23 <[hidden email]> wrote:

Dear Ryan,

1) Type of data: continuous 

2) Unit of measurement: (1) leverage: percentage; (2) directorship: integer

3) Skewness: (1) leverage: 1.94; (2) directorship: 1.42

4) Research question: (1) is the level of leverage affects company's decision to perform..... (2) is the number of directorship affects company's decision to perform.....

5) Zero values : (1) leverage: 19% of total sample ; (2) directorship: 15%

6) Data collection: secondary data from company annual reports

--- On Sun, 13/3/11, R B [via SPSSX Discussion] <[hidden email]> wrote:

From: R B [via SPSSX Discussion] <[hidden email]>
Subject: Re: how to handle absolute zero values?
To: "lcl23" <[hidden email]>
Date: Sunday, 13 March, 2011, 10:22 PM

Do you have two binary variables or two continuous variables? If the
latter, what units are they in? Approximately what do the shapes of
the distributions look like? Do they appear zero-inflated? What are
the research questions you're hoping to answer? How were the data
collected? Do you have a natural hierarchy in your data (e.g., persons
nested in families)?

Please provide more details.

Ryan

On Sat, Mar 12, 2011 at 8:25 PM, Bruce Weaver <[hidden email]> wrote:

> lcl23 wrote:
>>
>> There are 2 variables (ratio scale) in my data which have more than 15% of
>> absolute zero values, i.e. 0= no loan & 0= no asset. While entering the
>> data into SPSS, how should I deal with these absolute zero values? Should
>> I define all zeros as "missing values", which I think it might change the
>> meaning of these zeros? Or should I not doing anything and leave the data
>> as it is?
>>
>> I am going to analyse the data using logistic regression method. Thanks.
>>
>
>
> I don't understand where the problem is.  Unlike discriminant function
> analysis, logistic regression makes no assumptions about how explanatory
> variables are distributed.
>
>
>
> -----
> --
> 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/how-to-handle-absolute-zero-values-tp3451505p3545487.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

---

If you reply to this email, your message will be added to the discussion below:

http://spssx-discussion.1045642.n5.nabble.com/how-to-handle-absolute-zero-values-tp3451505p3555466.html

---

View this message in context: Re: how to handle absolute zero values?

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

[lcl23]

Looking at your questions, it seems that I need not do anything about those zeroes?

--- On Tue, 15/3/11, R B [via SPSSX Discussion] <[hidden email]> wrote:

From: R B [via SPSSX Discussion] <[hidden email]>
Subject: Re: how to handle absolute zero values?
To: "lcl23" <[hidden email]>
Date: Tuesday, 15 March, 2011, 8:58 AM

What led you to consider converting zeroes to missing values? What's the issue?

 

Ryan

On Sun, Mar 13, 2011 at 12:03 PM, lcl23 <[hidden email]> wrote:

Dear Ryan,

1) Type of data: continuous 

2) Unit of measurement: (1) leverage: percentage; (2) directorship: integer

3) Skewness: (1) leverage: 1.94; (2) directorship: 1.42

4) Research question: (1) is the level of leverage affects company's decision to perform..... (2) is the number of directorship affects company's decision to perform.....

5) Zero values : (1) leverage: 19% of total sample ; (2) directorship: 15%

6) Data collection: secondary data from company annual reports

--- On Sun, 13/3/11, R B [via SPSSX Discussion] <[hidden email]> wrote:

From: R B [via SPSSX Discussion] <[hidden email]>
Subject: Re: how to handle absolute zero values?
To: "lcl23" <[hidden email]>
Date: Sunday, 13 March, 2011, 10:22 PM

Do you have two binary variables or two continuous variables? If the
latter, what units are they in? Approximately what do the shapes of
the distributions look like? Do they appear zero-inflated? What are
the research questions you're hoping to answer? How were the data
collected? Do you have a natural hierarchy in your data (e.g., persons
nested in families)?

Please provide more details.

Ryan

On Sat, Mar 12, 2011 at 8:25 PM, Bruce Weaver <[hidden email]> wrote:

> lcl23 wrote:
>>
>> There are 2 variables (ratio scale) in my data which have more than 15% of
>> absolute zero values, i.e. 0= no loan & 0= no asset. While entering the
>> data into SPSS, how should I deal with these absolute zero values? Should
>> I define all zeros as "missing values", which I think it might change the
>> meaning of these zeros? Or should I not doing anything and leave the data
>> as it is?
>>
>> I am going to analyse the data using logistic regression method. Thanks.
>>
>
>
> I don't understand where the problem is.  Unlike discriminant function
> analysis, logistic regression makes no assumptions about how explanatory
> variables are distributed.
>
>
>
> -----
> --
> 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/how-to-handle-absolute-zero-values-tp3451505p3545487.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

---

If you reply to this email, your message will be added to the discussion below:

http://spssx-discussion.1045642.n5.nabble.com/how-to-handle-absolute-zero-values-tp3451505p3555466.html

---

View this message in context: Re: how to handle absolute zero values?

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

---

If you reply to this email, your message will be added to the discussion below:

http://spssx-discussion.1045642.n5.nabble.com/how-to-handle-absolute-zero-values-tp3451505p3635570.html

To unsubscribe from how to handle absolute zero values?, click here.

[lcl23]

Would it be alright if I don't do anything about those zeroes? 

--- On Mon, 14/3/11, Bruce Weaver [via SPSSX Discussion] <[hidden email]> wrote:

From: Bruce Weaver [via SPSSX Discussion] <[hidden email]>
Subject: Re: how to handle absolute zero values?
To: "lcl23" <[hidden email]>
Date: Monday, 14 March, 2011, 11:52 PM

Rich is describing Cohen & Cohen's "indicator" method for dealing with missing data.  It is no longer viewed favorably for situations where the data are truly missing (e.g., http://people.oregonstate.edu/~acock/growth-curves/working%20with%20missing%20values.pdf ).  But that is not the case here--the zeroes are legitimate values.  I don't recall ever reading anything that made the distinction between truly missing and the kind of gap present in this case.  But the indicator method might be all right here.

Bruce

Rich Ulrich wrote:

When "zero"  is nothing like being an equal-interval
extension of the other numbers observed, then the
proper way to model either score is to use two variables:
the new variable will be an indicator variable, 0/1  for
No/Yes, which indicates that "directors" (say)  exist.
With both variables going into the model, it does not actually
matter what value is used for the sort-of-missing score.

In some cases it is easier to note the effects if the
Missing is assigned the mean of the rest.

The same solution can work when a zero=no-event  exists
for data that otherwise seem appropriate for using the
log transform or a reciprocal:  Use the transform, and
set the missing to the mean of the rest.

--
Rich Ulrich

________________________________

> Date: Sun, 13 Mar 2011 09:03:41 -0700
> From: [hidden email]
> Subject: Re: how to handle absolute zero values?
> To: [hidden email]
>
> Dear Ryan,
>
> 1) Type of data: continuous
> 2) Unit of measurement: (1) leverage: percentage; (2) directorship: integer
> 3) Skewness: (1) leverage: 1.94; (2) directorship: 1.42
> 4) Research question: (1) is the level of leverage affects company's
> decision to perform..... (2) is the number of directorship affects
> company's decision to perform.....
> 5) Zero values : (1) leverage: 19% of total sample ; (2) directorship: 15%
> 6) Data collection: secondary data from company annual reports
>

[snip, previous]


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

---

If you reply to this email, your message will be added to the discussion below:

http://spssx-discussion.1045642.n5.nabble.com/how-to-handle-absolute-zero-values-tp3451505p3556569.html

To unsubscribe from how to handle absolute zero values?, click here.

[Rich Ulrich]

If the variable doesn't mean anything in the analysis,
then it doesn't matter what you do, as far as the outcome
of the analysis is concerned.

Since you don't have a lot of experience with analyses,
you should probably try three or four different things,
just to see what they do and how they differ, and whether
*any*  of them seem to give a meaningful result.

We don't know enough about the problem to say whether it
makes sense to use the original scoring (Can you argue that
the variable has "equal-intervals"?) or to use that scoring
except that you also drop the "zero" cases as missing; or
use zero and include a second, indicator variable for zero/
other; or use some other transformation.

--
Rich Ulrich
________________________________
> Date: Mon, 14 Mar 2011 20:17:31 -0700
> From: [hidden email]
> Subject: Re: how to handle absolute zero values?
> To: [hidden email]
>
> Would it be alright if I don't do anything about those zeroes?
>



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

[lcl23]

thanks, definitely will try all possible ways. 

--- On Wed, 16/3/11, Rich Ulrich [via SPSSX Discussion] <[hidden email]> wrote:

From: Rich Ulrich [via SPSSX Discussion] <[hidden email]>
Subject: Re: how to handle absolute zero values?
To: "lcl23" <[hidden email]>
Date: Wednesday, 16 March, 2011, 4:34 AM

If the variable doesn't mean anything in the analysis,
then it doesn't matter what you do, as far as the outcome
of the analysis is concerned.

Since you don't have a lot of experience with analyses,
you should probably try three or four different things,
just to see what they do and how they differ, and whether
*any*  of them seem to give a meaningful result.

We don't know enough about the problem to say whether it
makes sense to use the original scoring (Can you argue that
the variable has "equal-intervals"?) or to use that scoring
except that you also drop the "zero" cases as missing; or
use zero and include a second, indicator variable for zero/
other; or use some other transformation.

--
Rich Ulrich
________________________________
> Date: Mon, 14 Mar 2011 20:17:31 -0700
> From: [hidden email]
> Subject: Re: how to handle absolute zero values?
> To: [hidden email]
>
> Would it be alright if I don't do anything about those zeroes?
>



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

---

If you reply to this email, your message will be added to the discussion below:

http://spssx-discussion.1045642.n5.nabble.com/how-to-handle-absolute-zero-values-tp3451505p3719873.html

To unsubscribe from how to handle absolute zero values?, click here.