Date Variable

[Weinberg, Jerry]

I have a date variable: CERT_DATE  in the form dd-mmm-yyyy, I would like to use the following syntax:

 

IF   ANY (CERT_DATE,01-MAY-2012, 01-AUG-2012,01-DEC-2011) RETURNED_12RF='GOT DEGREE'.

EXECUTE.

 

However it doesn’t work, any suggestions??

 

 

Dr. Jerry Weinberg

Director of Institutional Research

St. Thomas University

305 474-6886

[Ruben Geert van den Berg]

Dear Jerry,

Please try the code below.

HTH,

Ruben

data list free/cert_date(a10).

begin data

'01.05.2012'

'02.05.2012'

'01.08.2012'

'02.08.2012'

end data.

alter type cert_date(edate8).

compute returned_12rf=ANY (CERT_DATE,date.dmy(01,05,2012),date.dmy(01,08,2012),date.dmy(01,12,2012)) .

EXECUTE./*Not needed but will show data in data view.*/

value labels returned_12rf 1'Got degree'0'FAILURE'.

---

Date: Wed, 31 Oct 2012 16:02:17 +0000
From: [hidden email]
Subject: Date Variable
To: [hidden email]

I have a date variable: CERT_DATE  in the form dd-mmm-yyyy, I would like to use the following syntax:

 

IF   ANY (CERT_DATE,01-MAY-2012, 01-AUG-2012,01-DEC-2011) RETURNED_12RF='GOT DEGREE'.

EXECUTE.

 

However it doesn’t work, any suggestions??

 

 

Dr. Jerry Weinberg

Director of Institutional Research

St. Thomas University

305 474-6886

[Richard Ristow]

At 12:02 PM 10/31/2012, Weinberg, Jerry wrote:

>I have a date variable: CERT_DATE  in the form dd-mmm-yyyy, I would
>like to use the following syntax:
>
>IF   ANY (CERT_DATE,01-MAY-2012, 01-AUG-2012,01-DEC-2011)
>      RETURNED_12RF='GOT DEGREE'.
>
>However it doesn't work, any suggestions??

Ruben van den Berg's looks like it should work, and likely has.

A little more on what's going on. You write, "I have a date variable:
CERT_DATE  in the form dd-mmm-yyyy". The thing is, you don't, not if
it's an SPSS date variable. Your variable may *display* as dd-mm-yyyy
(format DATE11), but its internal form is quite different, and is the
same for all SPSS date variables, however they display. (For
information: The internal form of a date variable is a number, the
number of seconds since midnight, October 14, 1582.)

Now, you're trying to compare this value with "01-MAY-2012",
"01-AUG-2012", "01-DEC-2011", treating those as constants whose
values are dates. Unfortunately, they aren't; in fact, regrettably,
SPSS has no way to write date-valued constants. Ruben took the usual
solution: write date-valued *expressions* instead:

IF   ANY (CERT_DATE,
           date.dmy(01,05,2012), date.dmy(01,08,2012),date.dmy(01,12,2012))
      RETURNED_12RF='GOT DEGREE'.

(I've rewritten Ruben's code a little, to match your original more closely.)

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

[Mark Lenel]

Hi,

 

Thanks for your message.  I am currently out of the office with limited access to emails.  I will be back in the office on Monday 5th November.

 

Should you require a faster response, please call the office [phone] and ask to speak to Crawford Christie, or email him on [hidden email].

 

Thanks!

 

Mark

Hello,

I've been trying to find the function for "Impute Missing Values" in the
SPSS software (20.0.0) for Mac OS. I found Add-Ons menu has this "SPSS
Missing Values" that seems to include "impute missing values", but I was
unable to locate where and how I can get this.

I got the SPSS software through my university (at a lot lower price).

Thanks for your help!

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

The Missing Value option is, well, an option.  You should find out whether they licensed it.  This is not related specifically to the Mac.


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




From:        "Onishi, Tamaki" <[hidden email]>
To:        [hidden email],
Date:        10/31/2012 08:09 PM
Subject:        [SPSSX-L] Mac version SPSS impute  missing value
Sent by:        "SPSSX(r) Discussion" <[hidden email]>

---

Hello,

I've been trying to find the function for "Impute Missing Values" in the
SPSS software (20.0.0) for Mac OS. I found Add-Ons menu has this "SPSS
Missing Values" that seems to include "impute missing values", but I was
unable to locate where and how I can get this.

I got the SPSS software through my university (at a lot lower price).

Thanks for your help!

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

[ViAnn Beadle]

not sure about site licenses or server licenses but shouldn't  the Show LICENSE show which options are licensed on a client machine?

On Wed, Oct 31, 2012 at 8:07 PM, Onishi, Tamaki <[hidden email]> wrote:

Hello,

I've been trying to find the function for "Impute Missing Values" in the
SPSS software (20.0.0) for Mac OS. I found Add-Ons menu has this "SPSS
Missing Values" that seems to include "impute missing values", but I was
unable to locate where and how I can get this.

I got the SPSS software through my university (at a lot lower price).

Thanks for your help!

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

--

My blog: lifeinnewmexico.wordpress.com

My pictures: http://picasaweb.google.com/vab88011

Twitter: vibeadle

Hello, 

(1) centering variables

Two professors of mine recently said it is not necessarily to center variables in the moderated regression. I was also reviewing some articles, and found the following sentences: 

Mean-centering not only reduces the covariance between x1 and x1x2, which is “good,” but it also reduces the variance of the exogenous variable x1x2, which is “bad.” For accurate measurement of the slope of the relationship, we need the exogenous variables to sweep out a large set of values; however, mean- centered (x1- x 1) (x2- x 2) has a smaller spread than x1x2. When both the improvement in collinearity and the deterioration of exogenous variable spread are considered, mean-centering provides no change in the accuracy with which the regression coefficients are estimated. The complete analysis of mean-centering shows that mean-centering neither helps nor hurts moderated regression.

This is my first time to have been using moderated regression and my original background is not quantitative field. So, I've been having a hard time in determining whether or not I should center variables. Could anybody let me know if I have better center variables, or in what situation I should center them, etc.? 

(2) using categorical variables 

I am using one variable --- the organization's legal status, nonprofit or for-profit — as independent variable. I created dummy variables for "For-proftit status" (or "nonprofit status", but use either one in models). One book said we should use only continuous variables for hierarchical regressions. Is this the case for moderated regression as well? If I still need to use the legal status in my model, how should I do it? 

Thank you so much, 

[Bruce Weaver]

(1)  The main reason for centering, IMO, is to make the coefficients more interpretable.  E.g., suppose your model contains, Age, years of education, and their product.  If you don't center, the constant gives the fitted value of Y when Age and Education are both 0; and the coefficients for Age and Education give you the change in Y-prime for a one unit increase in each of those variables while the other is equal to 0.  Zero is not a realistic value for Age, and probably not for Education either.  So the coefficients are not very easy to interpret.  But centering doesn't change the fit of the model (e.g., look at the R-squared value with and without centering).  Also, run your model with and without centering, and save the fitted values of Y (and residuals).  You'll get the same values either way.

(2) A dichotomous variable coded 1-0 can be treated exactly the same as a continuous variable.  

p.s. - Dividing variables by a constant can also help make coefficients easier to interpret.  E.g., I have sometimes centered Age on a value around the minimum for my sample, and then divided by 5 (or 10), because a one-year increment in age is too small to have any important impact on Y-prime.

HTH.

Onishi, Tamaki wrote

Hello,

(1) centering variables
Two professors of mine recently said it is not necessarily to center variables in the moderated regression. I was also reviewing some articles, and found the following sentences:


Mean-centering not only reduces the covariance between x1 and x1x2, which is “good,” but it also reduces the variance of the exogenous variable x1x2, which is “bad.” For accurate measurement of the slope of the relationship, we need the exogenous variables to sweep out a large set of values; however, mean- centered (x1- x 1) (x2- x 2) has a smaller spread than x1x2. When both the improvement in collinearity and the deterioration of exogenous variable spread are considered, mean-centering provides no change in the accuracy with which the regression coefficients are estimated. The complete analysis of mean-centering shows that mean-centering neither helps nor hurts moderated regression.


This is my first time to have been using moderated regression and my original background is not quantitative field. So, I've been having a hard time in determining whether or not I should center variables. Could anybody let me know if I have better center variables, or in what situation I should center them, etc.?


(2) using categorical variables

I am using one variable --- the organization's legal status, nonprofit or for-profit — as independent variable. I created dummy variables for "For-proftit status" (or "nonprofit status", but use either one in models). One book said we should use only continuous variables for hierarchical regressions. Is this the case for moderated regression as well? If I still need to use the legal status in my model, how should I do it?


Thank you so much,

[Poes, Matthew Joseph]

What you have found is essentially correct.  Mean Centering in moderated regression doesn’t really fix the problem it was supposed to fix and the net result is a no more accurate set of regression coefficients. 

 

To answer your questions, Mean centering is still often suggested as a means to make interpretation easier.  Beyond that, it isn’t necessary.  It’s easier to interpret because then the coefficients of the interaction are the difference from the mean.  Another thing, plot your interaction.

 

I’m not sure why you couldn’t use a dummy variable in a hierarchical regression.  If you can reference the book that indicated this, I might be able to make more sense of their claim.  In the example you have given, there is nothing I can see wrong with using them.  The Dummy variable still explains variance, it will still change the blocks r squared value if significant.

 

Matthew J Poes

Research Data Specialist

Center for Prevention Research and Development

University of Illinois

510 Devonshire Dr.

Champaign, IL 61820

Phone: [phone]

email: [hidden email]

 

 

From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Onishi, Tamaki
Sent: Thursday, November 01, 2012 12:31 AM
To: [hidden email]
Subject: Moderated regression - should center variable or not? & how to use categorical variable

 

Hello, 

 

(1) centering variables

Two professors of mine recently said it is not necessarily to center variables in the moderated regression. I was also reviewing some articles, and found the following sentences: 

 

Mean-centering not only reduces the covariance between x1 and x1x2, which is “good,” but it also reduces the variance of the exogenous variable x1x2, which is “bad.” For accurate measurement of the slope of the relationship, we need the exogenous variables to sweep out a large set of values; however, mean- centered (x1- x 1) (x2- x 2) has a smaller spread than x1x2.

When both the improvement in collinearity and the deterioration of exogenous variable spread are considered, mean-centering provides no change in the accuracy with which the regression coefficients are estimated. The complete analysis of mean-centering shows that mean-centering neither helps nor hurts moderated regression.

 

This is my first time to have been using moderated regression and my original background is not quantitative field. So, I've been having a hard time in determining whether or not I should center variables. Could anybody let me know if I have better center variables, or in what situation I should center them, etc.? 

 

(2) using categorical variables 

I am using one variable --- the organization's legal status, nonprofit or for-profit — as independent variable. I created dummy variables for "For-proftit status" (or "nonprofit status", but use either one in models). One book said we should use only continuous variables for hierarchical regressions. Is this the case for moderated regression as well? If I still need to use the legal status in my model, how should I do it? 

 

Thank you so much, 

 

Hello,

First, thank much for great advice from this Ml for my inquires.

(1) I believe I should put all IV and CV in one model. Exactly how (in
what order) should I put all DV and CV in addition to Interaction Term
when putting multiple DV? With one pair, I put 2 different IV ("org.
entrepreneurial orientation" and "nonprofit affiliation") in "Block 1 of
1" and then add Interaction Term for this pair to those 2 different IV in
the next block ("org. entrepreneurial orientation", "nonprofit
affiliation" and "interaction of entrepreneurial orientation and nonprofit
affiliation").

So, should I repeat this for each pair of IV? i.e.,

-- "org. entrepreneurial orientation" and "nonprofit affiliation") in
"Block 1 of 1"

-- "org. entrepreneurial orientation", "nonprofit affiliation" and
"interaction of entrepreneurial orientation and nonprofit affiliation" in
the next block
-- "org. entrepreneurial orientation" and "nonprofit affiliation" in the
next block -- starting with a new pair
-- "org. entrepreneurial orientation", "nonprofit affiliation"and
"interaction of entrepreneurial orientation and nonprofit affiliation"--
in the next block...


Of, should I put all DV pairs without interaction terms, before adding
interaction terms?


(2) Should I do the same for CV?

(3) I am very new to this, so forgive me if this is a too elementary
question. But what values (or numbers, etc.) should I include and explain
when I present those regression results? I usually look at R square and
Adjusted R square, F, VIF, tolerance, in addition to co-efficient
(Standardized and Unstandardized) and p-values. I also looked at Cronbach
alphas at the preliminary stage. I sometimes see researchers listing
results of correlation, etc. but am not sure why.

Thanks much.

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

[Richard Ristow]

At 10:01 AM 11/1/2012, Weinberg, Jerry wrote:

>   Thanks much for your reply, of course I didn't really want to do
> the work of recoding the date values from : 01-MAY-2012 to
> date.dmy(01,05,2012) in my syntax but if there is no way not to, of
> course I shall.

If you have a lot of those values, there's another way to do it that
is bulkier but may be easier changes for you to make. It involves
using function NUM, instead of DATE.DMY, and allows you to write
dates in something like the form you want.

Your original was

>IF   ANY (CERT_DATE,01-MAY-2012,01-AUG-2012,01-DEC-2011)
>      RETURNED_12RF='GOT DEGREE'.

You can rewrite that as

IF   ANY (CERT_DATE,NUM('01-MAY-2012',DATE11),NUM('01-AUG-2012',DATE11),
                     NUM('01-DEC-2011',DATE11))
      RETURNED_12RF='GOT DEGREE'.

which you may find more readable; and I was able to make the change
semi-automatically, with search-and-replace in a text editor. (Code is tested.)

================================
APPENDIX: Code for test run
(Ruben van den Berg's test data)
================================
data list free/cert_date(edate10).
begin data
01.05.2012
02.05.2012
01.08.2012
02.08.2012
end data.

STRING RETURNED_12RF(A10).

IF   ANY (CERT_DATE,NUM('01-MAY-2012',DATE11),NUM('01-AUG-2012',DATE11),
                     NUM('01-DEC-2011',DATE11))
      RETURNED_12RF='GOT DEGREE'.

LIST.

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

Perhaps the following?

IF   ANY (STRING(CERT_DATE,DATE11),'01-MAY-2012','01-AUG-2012','01-DEC-2011')
      RETURNED_12RF='GOT DEGREE'.
---------

>IF   ANY (CERT_DATE,01-MAY-2012,01-AUG-2012,01-DEC-2011)
>      RETURNED_12RF='GOT DEGREE'.

You can rewrite that as

IF   ANY (CERT_DATE,NUM('01-MAY-2012',DATE11),NUM('01-AUG-2012',DATE11),
                     NUM('01-DEC-2011',DATE11))
      RETURNED_12RF='GOT DEGREE'.

Richard Ristow wrote

At 10:01 AM 11/1/2012, Weinberg, Jerry wrote:

>   Thanks much for your reply, of course I didn't really want to do
> the work of recoding the date values from : 01-MAY-2012 to
> date.dmy(01,05,2012) in my syntax but if there is no way not to, of
> course I shall.

If you have a lot of those values, there's another way to do it that
is bulkier but may be easier changes for you to make. It involves
using function NUM, instead of DATE.DMY, and allows you to write
dates in something like the form you want.

Your original was

>IF   ANY (CERT_DATE,01-MAY-2012,01-AUG-2012,01-DEC-2011)
>      RETURNED_12RF='GOT DEGREE'.

You can rewrite that as

IF   ANY (CERT_DATE,NUM('01-MAY-2012',DATE11),NUM('01-AUG-2012',DATE11),
                     NUM('01-DEC-2011',DATE11))
      RETURNED_12RF='GOT DEGREE'.

which you may find more readable; and I was able to make the change
semi-automatically, with search-and-replace in a text editor. (Code is tested.)

================================
APPENDIX: Code for test run
(Ruben van den Berg's test data)
================================
data list free/cert_date(edate10).
begin data
01.05.2012
02.05.2012
01.08.2012
02.08.2012
end data.

STRING RETURNED_12RF(A10).

IF   ANY (CERT_DATE,NUM('01-MAY-2012',DATE11),NUM('01-AUG-2012',DATE11),
                     NUM('01-DEC-2011',DATE11))
      RETURNED_12RF='GOT DEGREE'.

LIST.

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

[Richard Ristow]

At 02:52 PM 11/3/2012, David Marso wrote:
>Perhaps the following?
>
>IF   ANY
>(STRING(CERT_DATE,DATE11),'01-MAY-2012','01-AUG-2012','01-DEC-2011')
>       RETURNED_12RF='GOT DEGREE'.

Excellent variation on the theme. Certainly more compact, likely runs
a little faster.

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

[Rich Ulrich]

(2) is simpler.  In mathematical conception, regression is just
a general computational method for ANOVA, whether it deals with
categorical factors (using dummy variables) or continuous measures.
A dichotomy counts as a "continuous variable" for most purposes.

So if your book seems to say "only continuous variables", I would say
that the author is talking about some special cases, or you are not
interpreting the advice correctly. 

(1) - Like Matthew and Bruce, I agree that "interpretation" is where
centering helps.  Unlike them, I emphasize the question, "Why are
you doing an analysis if you are not concerned with interpretation?"

I like the rule that you read an interpret the main effects *before*
you enter the interaction terms; if you are not looking at the coefficients
and tests for them in the fuller model, then you can trust today's computer
programs to preserve accuracy of computation... which seems to be the
point of the professor you cite.  (In the old days, pre-centering was needed
just to avoid the computational errors demonstrated by the Longley dataset.)

If I am a data analyst handing off results to people who are not statisticians,
centering helps me by eliminating questions that are not relevant.  If you
are a new analyst, it will help you avoid the same irrelevancies. 

I think I do not follow the technical complaint about "exogenous variable
spread."  I *think* this should have implications only when the two main
effects are highly correlated, and I still don't understand what the quoted
statement is getting at.  But when two predictors are highly correlated,
you do have special problems of interpretation, starting with the main
effects since they are not independent.  With highly correlated predictors,
I try to find an alternative to that model.  Does it need both?  Can it use
one of them, plus some sort of difference?

--
Rich Ulrich

---

Date: Thu, 1 Nov 2012 05:30:32 +0000
From: [hidden email]
Subject: Moderated regression - should center variable or not? & how to use categorical variable
To: [hidden email]

Hello, 

(1) centering variables

Two professors of mine recently said it is not necessarily to center variables in the moderated regression. I was also reviewing some articles, and found the following sentences: 

Mean-centering not only reduces the covariance between x1 and x1x2, which is “good,” but it also reduces the variance of the exogenous variable x1x2, which is “bad.” For accurate measurement of the slope of the relationship, we need the exogenous variables to sweep out a large set of values; however, mean- centered (x1- x 1) (x2- x 2) has a smaller spread than x1x2. When both the improvement in collinearity and the deterioration of exogenous variable spread are considered, mean-centering provides no change in the accuracy with which the regression coefficients are estimated. The complete analysis of mean-centering shows that mean-centering neither helps nor hurts moderated regression.

This is my first time to have been using moderated regression and my original background is not quantitative field. So, I've been having a hard time in determining whether or not I should center variables. Could anybody let me know if I have better center variables, or in what situation I should center them, etc.? 

(2) using categorical variables 

I am using one variable --- the organization's legal status, nonprofit or for-profit — as independent variable. I created dummy variables for "For-proftit status" (or "nonprofit status", but use either one in models). One book said we should use only continuous variables for hierarchical regressions. Is this the case for moderated regression as well? If I still need to use the legal status in my model, how should I do it? 

Thank you so much,