Dates: Subtracting number of months from a date
Locked
 
|
i have a variable called RepDate, which is in Date format. I have another variable called Age, which is a number to represetn the number of months since Issue Date. I want to calculate the Issue Date by subtracting from RepDate the number of months. Does anyone know hwo to do this in SPSS? I see a lot on subtracting one date from another but nto much on how to subtract a number of months from a date. Thanks so much in advance!
|
Re: Dates: Subtracting number of months from a date
|
|
You can use the DateSum command to add a specified number of units to a date.
The syntax looks like this... DATESUM(datevar, value, "unit", "method"). For this case, your syntax would look like this... COMPUTE newdate = DATESUM(repdate, -age, "months", 'closest'). Notice that the age is negative, this is because you are subtracting the age from the repdate and I am assuming that your age variable represents a positive number. -Steve 2010/1/14 jimjohn <[hidden email]> i have a variable called RepDate, which is in Date format. I have another |
Re: Dates: Subtracting number of months from a date
|
In reply to this post by jimjohn
----- Original Message -----
From: [hidden email]
To: [hidden email]
Sent: Friday, January 15, 2010 4:38 PM
Subject: RE: Dates: Subtracting number of months from a
date Dear
John, Please
see below it should work. Jacqueline COMMENT
as SPSS does all its date calculations in seconds, you need to convert Age into
seconds (as in brackets below below) then subtract from RepDate and then just
format the new IssueDate variable as date and time . COMPUTE
IssueDate=RepDate - (Age*30.417*24*60*60). FORMATS
IssueDate(EDATE10) . EXECUTE. That
should do it I have checked it with some data of my own.
Professor
Jacqueline Collier From: John F
Hall [mailto:[hidden email]] ----- Original
Message ----- From: [hidden email]
To: [hidden email]
Sent: Thursday, January
14, 2010 11:39 PM Subject: Dates: Subtracting
number of months from a date
This message has been checked for viruses but the contents of an attachment may still contain software viruses which could damage your computer system: you are advised to perform your own checks. Email communications with the University of Nottingham may be monitored as permitted by UK legislation.
|
Re: Dates: Subtracting number of months from a date
|
|
If you want calendar-correct results, you should use the DATESUM/DATEDIFF functions. If you want to use the more arithmetic approach, the date functions in the xdate and date groups and yrmoda will give more readable syntax and eliminate some sources of errors. The Date and Time wizard on the Transform menu can walk you through these calculations. HTH, Jon Peck SPSS, an IBM Company [hidden email] 312-651-3435
----- Original Message ----- From: Jacqueline Collier To: John F Hall Sent: Friday, January 15, 2010 4:38 PM Subject: RE: Dates: Subtracting number of months from a date Dear John, Please see below – it should work. Jacqueline
COMMENT as SPSS does all its date calculations in seconds, you need to convert Age into seconds (as in brackets below below) then subtract from RepDate and then just format the new IssueDate variable as date and time .
COMPUTE IssueDate=RepDate - (Age*30.417*24*60*60). FORMATS IssueDate(EDATE10) . EXECUTE.
That should do it – I have checked it with some data of my own.
Professor Jacqueline
Collier
From: John F Hall [mailto:johnfhall@...]
----- Original Message ----- From: jimjohn To: [hidden email] Sent: Thursday, January 14, 2010 11:39 PM Subject: Dates: Subtracting number of months from a date
This message has been checked for viruses but the contents of an attachment may still contain software viruses which could damage your computer system: you are advised to perform your own checks. Email communications with the University of Nottingham may be monitored as permitted by UK legislation. ----- Original Message -----
|
different results for mixed and GLM
|
|
In reply to this post by smarden1@gmail.com
Hello,
I want to run an analysis with a continuous
dependent variable, a continuous independent variable, a dichotomous
independent variable, and the interaction between both independent variables.
I've run the analysis using both GLM (univariate) and mixed
models, and the results turn out to be different. Aren't the results
supposed to be identical or do Mixed and GLM estimate different
models?
Joost van Ginkel
Joost R. Van Ginkel,
PhD ********************************************************************** This email and any files transmitted with it are confidential and intended solely for the use of the individual or entity to whom they are addressed. If you have received this email in error please notify the system manager. **********************************************************************
|
Re: different results for mixed and GLM
|
Do you have nested levels of cases? Can you cobble together a small dataset with DATA LIST a few cases that has a structure similar to your data as well as the syntax for the two procedures that is giving results that are causing you a problem? Art Kendall Social Research Consultants On 1/16/2010 7:08 AM, Ginkel, Joost van wrote: ===================== 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 |
Re: different results for mixed and GLM
|
|
Dear Art,
I just found out that the results do not differ. I
got confused because I mixed up the results of
the F-tests with the results of the t-tests of the parameters estimates. Results
of the F-tests do not differ for mixed and GLM, and neither do the results of
the t-tests. However, the results of the F-tests do differ from the results of
the t-tests for the continuous variable. So this raises a new question: in
an analysis with a dichotomous independent variable and a continuous independent
variable, aren't the F-values simply the squared values of the t-tests? In this
case I would expect the results of the F-test and the t-tests to be identical.
However, this is not the case. How come? And if the results are indeed
supposed to be different, then what does the one test tell us and what does the
other one tell us?
Best
regards,
Joost van
Ginkel Joost R. Van Ginkel,
PhD From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Art Kendall xxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxx Sent: 16 January 2010 21:14 To: [hidden email] Subject: Re: different results for mixed and GLM Do you have nested levels of cases? Can you cobble together a small dataset with DATA LIST a few cases that has a structure similar to your data as well as the syntax for the two procedures that is giving results that are causing you a problem? Art Kendall Social Research Consultants On 1/16/2010 7:08 AM, Ginkel, Joost van wrote: ===================== 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 |
Re: different results for mixed and GLM
|
|
In reply to this post by Joost van Ginkel
If you have missing data, that could account for it. GLM omits
whole cases if any variable is missing, but Mixed does not. Garry Business Analytic Ltd From: SPSSX(r) Discussion
[mailto:[hidden email]] On Behalf Of Ginkel, Joost van Hello, I want to run an analysis with a continuous dependent
variable, a continuous independent variable, a dichotomous independent
variable, and the interaction between both independent variables. I've run
the analysis using both GLM (univariate) and mixed models, and the results
turn out to be different. Aren't the results supposed to be identical
or do Mixed and GLM estimate different models? Joost van Ginkel Joost R. Van
Ginkel, PhD ********************************************************************** This email
and any files transmitted with it are confidential and intended
solely for the use of the individual or entity to whom they are
addressed. If you have received this email in error please notify the system
manager. ********************************************************************** __________ Information from ESET NOD32 Antivirus, version of virus signature database 4782 (20100118) __________ The message was checked by ESET NOD32 Antivirus. http://www.eset.com |
Re: different results for mixed and GLM
|
|
In reply to this post by Joost van Ginkel
Dear Joost van Ginkel,
I am new to the world of statistics, but as far as I know between GLM and Mixed .... the idea is the same, but the way and considerations are different. Can help you: UGRINOWITSCH, CARLOS; FELLINGHAM, GILBERT W.; RICARD, MARK D. Limitations of Ordinary Least Squares Models in Analyzing Repeated Measures Data. Medicine & Science in Sports & Exercise: December 2004 - Volume 36 - Issue 12 - pp 2144-2148 Good Luck, Luciano Basso University of São Paulo Laboratory of Motor Behavior São Paulo - Brazil ===================== 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 |
Re: different results for mixed and GLM
|
|
In reply to this post by Joost van Ginkel
Joost,
Please post your command syntax and the significance test results. I'm assuming the difference you seeing is not 'decimal dust', i.e., 1.998**2 vs 4.00, so I think it will be helpful to all to see details. Gene Maguin Dear Art, I just found out that the results do not differ. I got confused because I mixed up the results of the F-tests with the results of the t-tests of the parameters estimates. Results of the F-tests do not differ for mixed and GLM, and neither do the results of the t-tests. However, the results of the F-tests do differ from the results of the t-tests for the continuous variable. So this raises a new question: in an analysis with a dichotomous independent variable and a continuous independent variable, aren't the F-values simply the squared values of the t-tests? In this case I would expect the results of the F-test and the t-tests to be identical. However, this is not the case. How come? And if the results are indeed supposed to be different, then what does the one test tell us and what does the other one tell us? Best regards, Joost van Ginkel Joost R. Van Ginkel, PhD Leiden University Faculty of Social and Behavioural Sciences Data Theory Group PO Box 9555 2300 RB Leiden The Netherlands Tel: +31-(0)71-527 3620 Fax: +31-(0)71-527 1721 ________________________________ From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Art Kendall xxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxx Sent: 16 January 2010 21:14 To: [hidden email] Subject: Re: different results for mixed and GLM Please post a description of your data set. Do you have nested levels of cases? Can you cobble together a small dataset with DATA LIST a few cases that has a structure similar to your data as well as the syntax for the two procedures that is giving results that are causing you a problem? Art Kendall Social Research Consultants On 1/16/2010 7:08 AM, Ginkel, Joost van wrote: Hello, I want to run an analysis with a continuous dependent variable, a continuous independent variable, a dichotomous independent variable, and the interaction between both independent variables. I've run the analysis using both GLM (univariate) and mixed models, and the results turn out to be different. Aren't the results supposed to be identical or do Mixed and GLM estimate different models? Joost van Ginkel Joost R. Van Ginkel, PhD Leiden University Faculty of Social and Behavioural Sciences Data Theory Group PO Box 9555 2300 RB Leiden The Netherlands Tel: +31-(0)71-527 3620 Fax: +31-(0)71-527 1721 ********************************************************************** This email and any files transmitted with it are confidential and intended solely for the use of the individual or entity to whom they are addressed. If you have received this email in error please notify the system manager. ********************************************************************** ===================== 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 |
missing values in MIXED and GLM [was Re: different results for mixed and GLM]
|
|
In reply to this post by Garry Gelade
To expand a bit on the below: MIXED, like GLM, omits whole rows if any variable in the analysis has invalid values**; however, repeated measurements in MIXED are spread out over multiple rows, so MIXED will use the rows with valid values for a given subject. By contrast, repeated measurements in GLM are spread out over multiple variables and each subject occupies a single row, so if any variable for a given subject is invalid, all observations for that subject are thrown out. Alex ** MIXED does have the additional option of treating user-missing values of categorical variables as valid, so it is possible in a simple GLM Univariate model that user-missing values could cause different results between the two procedures.
If you have missing data, that could account for it. GLM omits whole cases if any variable is missing, but Mixed does not. Garry Business Analytic Ltd From: SPSSX(r) Discussion [[hidden email]] On Behalf Of Ginkel, Joost van Sent: 16 January 2010 12:09 To: [hidden email] Subject: different results for mixed and GLM Hello, I want to run an analysis with a continuous dependent variable, a continuous independent variable, a dichotomous independent variable, and the interaction between both independent variables. I've run the analysis using both GLM (univariate) and mixed models, and the results turn out to be different. Aren't the results supposed to be identical or do Mixed and GLM estimate different models? Joost van Ginkel Joost R. Van Ginkel, PhD
|
Re: different results for mixed and GLM
|
|
In reply to this post by Maguin, Eugene
Dear Gene,
Please, try this: Data list free/v1 v2 v3 . begin data 1 9.70068493150685 31 0 9.17214611872146 24 1 9.22808219178082 26 0 9.69246575342466 35 0 9.73082191780822 32 0 10.736301369863 23 1 9.72808219178082 27 0 9.53401826484018 33 1 9.06689497716895 28 0 9.53675799086758 32 1 9.71986301369863 31 0 9.69246575342466 33 0 9.29223744292237 31 0 10.9057077625571 33 end data. FORMAT V1 (F1.0) . MIXED v3 BY v1 WITH v2 /CRITERIA=CIN(95) MXITER(100) MXSTEP(5) SCORING(1) SINGULAR(0.000000000001) HCONVERGE(0, ABSOLUTE) LCONVERGE(0, ABSOLUTE) PCONVERGE(0.000001, ABSOLUTE) /FIXED=v1 v2 v1*v2 | SSTYPE(3) /METHOD=REML /PRINT=SOLUTION. UNIANOVA v3 BY v1 WITH v2 /METHOD=SSTYPE(3) /INTERCEPT=INCLUDE /PRINT=PARAMETER /CRITERIA=ALPHA(.05) /DESIGN=v2 v1 v1*v2. You'll see that the results for Mixed and GLM are the same but that the results of the t-tests and F-tests differ for the intercept and the continuous variable V2. So what I don't understand is: how comes it that the t- and F-tests give different p-values here? Best regards, Joost van Ginkel Joost R. Van Ginkel, PhD Leiden University Faculty of Social and Behavioural Sciences Data Theory Group PO Box 9555 2300 RB Leiden The Netherlands Tel: +31-(0)71-527 3620 Fax: +31-(0)71-527 1721 -----Original Message----- From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Gene Maguin Sent: 18 January 2010 16:34 To: [hidden email] Subject: Re: different results for mixed and GLM Joost, Please post your command syntax and the significance test results. I'm assuming the difference you seeing is not 'decimal dust', i.e., 1.998**2 vs 4.00, so I think it will be helpful to all to see details. Gene Maguin Dear Art, I just found out that the results do not differ. I got confused because I mixed up the results of the F-tests with the results of the t-tests of the parameters estimates. Results of the F-tests do not differ for mixed and GLM, and neither do the results of the t-tests. However, the results of the F-tests do differ from the results of the t-tests for the continuous variable. So this raises a new question: in an analysis with a dichotomous independent variable and a continuous independent variable, aren't the F-values simply the squared values of the t-tests? In this case I would expect the results of the F-test and the t-tests to be identical. However, this is not the case. How come? And if the results are indeed supposed to be different, then what does the one test tell us and what does the other one tell us? Best regards, Joost van Ginkel Joost R. Van Ginkel, PhD Leiden University Faculty of Social and Behavioural Sciences Data Theory Group PO Box 9555 2300 RB Leiden The Netherlands Tel: +31-(0)71-527 3620 Fax: +31-(0)71-527 1721 ________________________________ From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Art Kendall xxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxx Sent: 16 January 2010 21:14 To: [hidden email] Subject: Re: different results for mixed and GLM Please post a description of your data set. Do you have nested levels of cases? Can you cobble together a small dataset with DATA LIST a few cases that has a structure similar to your data as well as the syntax for the two procedures that is giving results that are causing you a problem? Art Kendall Social Research Consultants On 1/16/2010 7:08 AM, Ginkel, Joost van wrote: Hello, I want to run an analysis with a continuous dependent variable, a continuous independent variable, a dichotomous independent variable, and the interaction between both independent variables. I've run the analysis using both GLM (univariate) and mixed models, and the results turn out to be different. Aren't the results supposed to be identical or do Mixed and GLM estimate different models? Joost van Ginkel Joost R. Van Ginkel, PhD Leiden University Faculty of Social and Behavioural Sciences Data Theory Group PO Box 9555 2300 RB Leiden The Netherlands Tel: +31-(0)71-527 3620 Fax: +31-(0)71-527 1721 ********************************************************************** This email and any files transmitted with it are confidential and intended solely for the use of the individual or entity to whom they are addressed. If you have received this email in error please notify the system manager. ********************************************************************** ===================== 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 |
Re: different results for mixed and GLM
|
|
Joost,
I don't know the answer and I'd be interested to hear about this from someone that does. Specifically, in your model, v3 by v1 with v2, the square root of F equals the t value for the v1 and the v1*v2 effects but not for the v2 and intercept effects. Specifically, Effect F-value sqrt(F) t-value ----------------------------------- V1*v2 0.530 0.728 -0.728 V1 0.580 0.7615 0.762 V2 0.234 0.483 0.646 Int 0.213 0.462 -0.161 Gene Maguin ===================== 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 |
Re: different results for mixed and GLM
Administrator
|
In reply to this post by Joost van Ginkel
And to make it even more interesting, note that V1 is binary (coded 0-1), so treating it as continuous rather than categorical should not change the results (apart from the sign on a coefficient, perhaps). But try this: MIXED v3 WITH v1 v2 /CRITERIA=CIN(95) MXITER(100) MXSTEP(5) SCORING(1) SINGULAR(0.000000000001) HCONVERGE(0, ABSOLUTE) LCONVERGE(0, ABSOLUTE) PCONVERGE(0.000001, ABSOLUTE) /FIXED=v1 v2 v1*v2 | SSTYPE(3) /METHOD=REML /PRINT=SOLUTION. I get the same deviance as for the model with "BY V1", but there are other differences that seem a bit odd. I wonder if it has to do with centering of variables? And of course, one could also run the model via REGRESSION, like this: compute v1v2 = v1*v2. REGRESSION /STATISTICS COEFF OUTS CI(95) R ANOVA /DEPENDENT v3 /METHOD=TEST (v1) (v2) (v1v2) . Cheers, Bruce
--
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/). |
Re: different results for mixed and GLM
|
|
The reason is that different things are being tested:
For the F Table, the intercept is testing that the sum of the two intercepts = 0, and v2 is testing that the sum of the slopes = 0. While in the t - table, the intercept is testing that the second intercept = 0, and v2 is testing that the second slope = 0. Mike F. Michael Speed Professor Director of Online Learning Department of Statistics Associate Dean for Technology Mediated Instruction College of Science TAMU 979-845-3182 -----Original Message----- From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Bruce Weaver Sent: Tuesday, January 19, 2010 5:11 PM To: [hidden email] Subject: Re: different results for mixed and GLM Ginkel, Joost van wrote: > > Dear Gene, > > Please, try this: > > Data list free/v1 v2 v3 . > begin data > 1 9.70068493150685 31 > 0 9.17214611872146 24 > 1 9.22808219178082 26 > 0 9.69246575342466 35 > 0 9.73082191780822 32 > 0 10.736301369863 23 > 1 9.72808219178082 27 > 0 9.53401826484018 33 > 1 9.06689497716895 28 > 0 9.53675799086758 32 > 1 9.71986301369863 31 > 0 9.69246575342466 33 > 0 9.29223744292237 31 > 0 10.9057077625571 33 > end data. > FORMAT V1 (F1.0) . > > > MIXED v3 BY v1 WITH v2 > /CRITERIA=CIN(95) MXITER(100) MXSTEP(5) SCORING(1) > SINGULAR(0.000000000001) HCONVERGE(0, > ABSOLUTE) LCONVERGE(0, ABSOLUTE) PCONVERGE(0.000001, ABSOLUTE) > /FIXED=v1 v2 v1*v2 | SSTYPE(3) > /METHOD=REML > /PRINT=SOLUTION. > > UNIANOVA v3 BY v1 WITH v2 > /METHOD=SSTYPE(3) > /INTERCEPT=INCLUDE > /PRINT=PARAMETER > /CRITERIA=ALPHA(.05) > /DESIGN=v2 v1 v1*v2. > > You'll see that the results for Mixed and GLM are the same but that the > results of the t-tests and F-tests differ for the intercept and the > continuous variable V2. So what I don't understand is: how comes it that > the t- and F-tests give different p-values here? > > Best regards, > > Joost van Ginkel > > And to make it even more interesting, note that V1 is binary (coded 0-1), so treating it as continuous rather than categorical should not change the results (apart from the sign on a coefficient, perhaps). But try this: MIXED v3 WITH v1 v2 /CRITERIA=CIN(95) MXITER(100) MXSTEP(5) SCORING(1) SINGULAR(0.000000000001) HCONVERGE(0, ABSOLUTE) LCONVERGE(0, ABSOLUTE) PCONVERGE(0.000001, ABSOLUTE) /FIXED=v1 v2 v1*v2 | SSTYPE(3) /METHOD=REML /PRINT=SOLUTION. I get the same deviance as for the model with "BY V1", but there are other differences that seem a bit odd. I wonder if it has to do with centering of variables? And of course, one could also run the model via REGRESSION, like this: compute v1v2 = v1*v2. REGRESSION /STATISTICS COEFF OUTS CI(95) R ANOVA /DEPENDENT v3 /METHOD=TEST (v1) (v2) (v1v2) . Cheers, Bruce ----- -- 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://old.nabble.com/Dates%3A-Subtracting-number-of-months-from-a-date-tp27169326p27234394.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 |
Re: different results for mixed and GLM
|
|
Dear Mike,
Thank you very much for your answer (and of course all the others who responded to my message). This explains a lot! Best regards, Joost van Ginkel Joost R. Van Ginkel, PhD Leiden University Faculty of Social and Behavioural Sciences Data Theory Group PO Box 9555 2300 RB Leiden The Netherlands Tel: +31-(0)71-527 3620 Fax: +31-(0)71-527 1721 -----Original Message----- From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Mike Speed Sent: 20 January 2010 00:38 To: [hidden email] Subject: Re: different results for mixed and GLM The reason is that different things are being tested: For the F Table, the intercept is testing that the sum of the two intercepts = 0, and v2 is testing that the sum of the slopes = 0. While in the t - table, the intercept is testing that the second intercept = 0, and v2 is testing that the second slope = 0. Mike F. Michael Speed Professor Director of Online Learning Department of Statistics Associate Dean for Technology Mediated Instruction College of Science TAMU 979-845-3182 -----Original Message----- From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Bruce Weaver Sent: Tuesday, January 19, 2010 5:11 PM To: [hidden email] Subject: Re: different results for mixed and GLM Ginkel, Joost van wrote: > > Dear Gene, > > Please, try this: > > Data list free/v1 v2 v3 . > begin data > 1 9.70068493150685 31 > 0 9.17214611872146 24 > 1 9.22808219178082 26 > 0 9.69246575342466 35 > 0 9.73082191780822 32 > 0 10.736301369863 23 > 1 9.72808219178082 27 > 0 9.53401826484018 33 > 1 9.06689497716895 28 > 0 9.53675799086758 32 > 1 9.71986301369863 31 > 0 9.69246575342466 33 > 0 9.29223744292237 31 > 0 10.9057077625571 33 > end data. > FORMAT V1 (F1.0) . > > > MIXED v3 BY v1 WITH v2 > /CRITERIA=CIN(95) MXITER(100) MXSTEP(5) SCORING(1) > SINGULAR(0.000000000001) HCONVERGE(0, > ABSOLUTE) LCONVERGE(0, ABSOLUTE) PCONVERGE(0.000001, ABSOLUTE) > /FIXED=v1 v2 v1*v2 | SSTYPE(3) > /METHOD=REML > /PRINT=SOLUTION. > > UNIANOVA v3 BY v1 WITH v2 > /METHOD=SSTYPE(3) > /INTERCEPT=INCLUDE > /PRINT=PARAMETER > /CRITERIA=ALPHA(.05) > /DESIGN=v2 v1 v1*v2. > > You'll see that the results for Mixed and GLM are the same but that > the results of the t-tests and F-tests differ for the intercept and > the continuous variable V2. So what I don't understand is: how comes > it that the t- and F-tests give different p-values here? > > Best regards, > > Joost van Ginkel > > And to make it even more interesting, note that V1 is binary (coded 0-1), so treating it as continuous rather than categorical should not change the results (apart from the sign on a coefficient, perhaps). But try this: MIXED v3 WITH v1 v2 /CRITERIA=CIN(95) MXITER(100) MXSTEP(5) SCORING(1) SINGULAR(0.000000000001) HCONVERGE(0, ABSOLUTE) LCONVERGE(0, ABSOLUTE) PCONVERGE(0.000001, ABSOLUTE) /FIXED=v1 v2 v1*v2 | SSTYPE(3) /METHOD=REML /PRINT=SOLUTION. I get the same deviance as for the model with "BY V1", but there are other differences that seem a bit odd. I wonder if it has to do with centering of variables? And of course, one could also run the model via REGRESSION, like this: compute v1v2 = v1*v2. REGRESSION /STATISTICS COEFF OUTS CI(95) R ANOVA /DEPENDENT v3 /METHOD=TEST (v1) (v2) (v1v2) . Cheers, Bruce ----- -- 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://old.nabble.com/Dates%3A-Subtracting-number-of-months-from-a-date- tp27169326p27234394.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 ********************************************************************** This email and any files transmitted with it are confidential and intended solely for the use of the individual or entity to whom they are addressed. If you have received this email in error please notify the system manager. ********************************************************************** ===================== 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 |
| Free forum by Nabble | Edit this page |