SPSSX Discussion

Gender as Covariate in ANCOVA, Schools/Teachers as Dummy Variables, ANCOVA v. Regression?

Classic

List

Threaded

20 messages Options

Justin Meyer-3

Gender as Covariate in ANCOVA, Schools/Teachers as Dummy Variables, ANCOVA v. Regression?

Hi SPSSlisters,

I am trying to determine if a reading program has an effect on students'
attitudes towards reading. To do this, a reading attitude survey was
administered to 125 students in a treatment group (used the reading
program) and 96 students in a control group (did not use the reading
program) (this is the IV). The attitude survey results in a score (the
DV) that can range from 20 to 80 points. Because females scored higher
than males on the survey, I felt that it was necessary to incorporate
gender in the analysis. I ran an ANCOVA with gender as a covariate and
found that both gender (p=.01) and treatment (p=.029) are significant
predictors of attitude score. I have been asked to investigate two other
factors: school (there are three schools, with both treatment and
control students at each) and teacher (there are seven treatment
teachers and five control). Below are the questions I have:

1. Does it make sense to enter gender as a covariate or should it
have been entered as a fixed factor? More generally, how do I know if
something should be entered as a covariate or a fixed factor?

2. Can I enter school as a fixed factor (one variable with school
A=1, school B=2, school C=3) or should I code school as two dummy
variables (are these then entered as fixed factors)?

3. How do I represent the teachers? Dummy variables or some other
method?

4. Should all this be done using a regression instead of ANCOVA?

Based on what I know, a hierarchical analysis would be the best way to
incorporate teacher and school, but I don't think I have a sufficient
number of teachers or schools for this.

I would appreciate any suggestions you might have. I will likely have
more SPSS-specific questions once I figure out what I need to do.

Thank you,

____________________________________

Justin Meyer

Researcher

Rowland Reading Foundation

www.rowlandreading.org <http://www.rowlandreading.org/>

____________________________________

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

Marta Garcia-Granero

Re: Gender as Covariate in ANCOVA, Schools/Teachers as Dummy Variables, ANCOVA v. Regression?

Justin Meyer wrote:

> I am trying to determine if a reading program has an effect on students'
> attitudes towards reading. To do this, a reading attitude survey was
> administered to 125 students in a treatment group (used the reading
> program) and 96 students in a control group (did not use the reading
> program) (this is the IV). The attitude survey results in a score (the
> DV) that can range from 20 to 80 points. Because females scored higher
> than males on the survey, I felt that it was necessary to incorporate
> gender in the analysis. I ran an ANCOVA with gender as a covariate and
> found that both gender (p=.01) and treatment (p=.029) are significant
> predictors of attitude score. I have been asked to investigate two other
> factors: school (there are three schools, with both treatment and
> control students at each) and teacher (there are seven treatment
> teachers and five control). Below are the questions I have:
>
>
>
> 1. Does it make sense to enter gender as a covariate or should it
> have been entered as a fixed factor? More generally, how do I know if
> something should be entered as a covariate or a fixed factor?
>

Covariates are quantitative variables (usually random "noise" to be
controlled), therefore, Gender must enter the model as a fixed factor.
> 2. Can I enter school as a fixed factor (one variable with school
> A=1, school B=2, school C=3) or should I code school as two dummy
> variables (are these then entered as fixed factors)?
>

Fixed factor directly, no need to dummy code them.
> 3. How do I represent the teachers? Dummy variables or some other
> method?
>

I would enter them as random factors, nested within treatments.
> 4. Should all this be done using a regression instead of ANCOVA?
>

I think ANOVA should be OK

HTH,
Marta García-Granero

I want to thank everyone who wrote to me with warm words of comfort
concerning the terrorist attack at my University (injuring more than 20
students&workers). To the serbian guy that was rather revengefully glad
that it had happened because in 1999, allegedly, "Spanish bombers
attacked Serbia and killed children" (that being absolutely false,
Spanish troops were there under UN flag with humanitarian tasks, to
prevent civilians from being slaughtered by exalted people from both
sides), I can only say that he/she (there wasn't a signature on the
venomous message I got) reminds me a lot of the people who attacked, for
the 6th time in the last 30 years, my University: ignorant and full of
hate people. Knowledge makes people better, he/she should get informed
before getting to those sickly conclusions.

--
For miscellaneous statistical stuff, visit:
http://gjyp.nl/marta/

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

Zdaniuk, Bozena-2

Re: Gender as Covariate in ANCOVA, Schools/Teachers as Dummy Variables, ANCOVA v. Regression?

I have brought up (see below) a part of Marta's message I have a question about: I don't think you can designate factors as nested in the regular GLM ANOVA spss analysis, am I correct? I think one must use the Gneralized (as opposed to General) LM analysis. If I am wrong, please let me know how one can do nesting with the General LM.
thanks.
Bozena Zdaniuk

Fixed factor directly, no need to dummy code them.
> 3. How do I represent the teachers? Dummy variables or some other
> method?
>
I would enter them as random factors, nested within treatments.
> 4. Should all this be done using a regression instead of ANCOVA?
>
I think ANOVA should be OK
HTH,
Marta García-Granero

________________________________________
From: SPSSX(r) Discussion [[hidden email]] On Behalf Of Marta García-Granero [[hidden email]]
Sent: Thursday, October 30, 2008 2:56 PM
To: [hidden email]
Subject: Re: Gender as Covariate in ANCOVA, Schools/Teachers as Dummy Variables, ANCOVA v. Regression?

Justin Meyer wrote:

Bob Schacht-3

Re: Gender as Covariate in ANCOVA, Schools/Teachers as Dummy Variables, ANCOVA v. Regression?

In reply to this post by Marta Garcia-Granero

At 08:56 AM 10/30/2008, Marta García-Granero wrote:
>Justin Meyer wrote:
>I am trying to determine if a reading program has an effect on students'
>attitudes towards reading. . . .

[snip]

>>4. Should all this be done using a regression instead of ANCOVA?
>
>I think ANOVA should be OK

For the problem as stated, as Marta wrote, an ANOVA (sic!) should be
sufficient. The only reason that I can see for doing a regression instead
would be if there was some need for the regression equation. Except for
computing the regression coefficient(s) and constant, don't these two
procedures overlap considerably in their computations?

In some ways, ANOVA would be *more* appropriate than regression, because
regression (at least in its classical form) is not designed for use with
categorical variables, whereas ANOVA is built to handle a mix of variables.
So ISTM, ANOVA would be a more appropriate model for the analysis.

Bob Schacht

Robert M. Schacht, Ph.D. <[hidden email]>
Pacific Basin Rehabilitation Research & Training Center
1268 Young Street, Suite #204
Research Center, University of Hawaii
Honolulu, HI 96814

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

Justin Meyer-3

Re: Gender as Covariate in ANCOVA, Schools/Teachers as Dummy Variables, ANCOVA v. Regression?

In reply to this post by Marta Garcia-Granero

Thank you very much for responding to my questions. I really appreciate it. I too am very sorry for the attack on your university. It's extremely troubling that terrorist attacks seem to be becoming more common rather than less. Attacking innocent people is never right.

Regarding your response to question 3 below, could you please describe how to nest the teachers within treament? Would I create one new variable, with teacher A=1, teacher B=2, etc.?

Thank you,
____________________________________

Justin Meyer

Rowland Reading Foundation

____________________________________

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Marta García-Granero
Sent: Thursday, October 30, 2008 1:57 PM
To: [hidden email]
Subject: Re: Gender as Covariate in ANCOVA, Schools/Teachers as Dummy Variables, ANCOVA v. Regression?

Justin Meyer wrote:

Marta Garcia-Granero

Re: Gender as Covariate in ANCOVA, Schools/Teachers as Dummy Variables, ANCOVA v. Regression?

In reply to this post by Zdaniuk, Bozena-2

Zdaniuk, Bozena wrote:
> I have brought up (see below) a part of Marta's message I have a question about: I don't think you can designate factors as nested in the regular GLM ANOVA spss analysis, am I correct? I think one must use the Gneralized (as opposed to General) LM analysis. If I am wrong, please let me know how one can do nesting with the General LM.
> thanks.
>
You can use UNIANOVA, but not with the point&click interface, or you can
use MIXED:

* Sample dataset, 4 cages, with 4 rats each, are randomly assigned to 2
different ionizations 8negative 6 positive *.
DATA LIST LIST/ionization cage activity (3 F8).
BEGIN DATA
1 1 3
1 1 6
1 1 3
1 1 3
1 2 1
1 2 2
1 2 2
1 2 2
1 3 5
1 3 6
1 3 5
1 3 6
1 4 2
1 4 3
1 4 4
1 4 3
2 5 7
2 5 8
2 5 7
2 5 6
2 6 4
2 6 5
2 6 4
2 6 3
2 7 7
2 7 8
2 7 9
2 7 8
2 8 10
2 8 10
2 8 9
2 8 11
END DATA.

VALUE LABEL ionization 1'Negative' 2'Positive'.

UNIANOVA
activity BY ionization cage
/RANDOM = cage
/METHOD = SSTYPE(1)
/DESIGN = ionization cage WITHIN ionization .

MIXED
activity BY ionization cage
/FIXED = ionization | SSTYPE(1)
/METHOD = REML
/RANDOM cage(ionization) | COVTYPE(VC) .

HTH,
Marta

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

Zdaniuk, Bozena-2

Re: Gender as Covariate in ANCOVA, Schools/Teachers as Dummy Variables, ANCOVA v. Regression?

In reply to this post by Justin Meyer-3

Yes, you need a variable 'teacher' with a number denoting a teacher for each case (i am assuming your cases/rows of data are students) .
bozena
________________________________________
From: SPSSX(r) Discussion [[hidden email]] On Behalf Of Justin Meyer [[hidden email]]
Sent: Thursday, October 30, 2008 3:23 PM
To: [hidden email]
Subject: Re: Gender as Covariate in ANCOVA, Schools/Teachers as Dummy Variables, ANCOVA v. Regression?

Thank you very much for responding to my questions. I really appreciate it. I too am very sorry for the attack on your university. It's extremely troubling that terrorist attacks seem to be becoming more common rather than less. Attacking innocent people is never right.

Regarding your response to question 3 below, could you please describe how to nest the teachers within treament? Would I create one new variable, with teacher A=1, teacher B=2, etc.?

Thank you,
____________________________________

Justin Meyer

Rowland Reading Foundation

____________________________________

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Marta García-Granero
Sent: Thursday, October 30, 2008 1:57 PM
To: [hidden email]
Subject: Re: Gender as Covariate in ANCOVA, Schools/Teachers as Dummy Variables, ANCOVA v. Regression?

Justin Meyer wrote:

Justin Meyer-3

Re: Gender as Covariate in ANCOVA, Schools/Teachers as Dummy Variables, ANCOVA v. Regression?

In reply to this post by Marta Garcia-Granero

Thanks to everyone who responded to my questions yesterday and thanks to Marta for the syntax. I was able to continue my analysis with your help. A few more questions:

1. I ran an ANOVA with attitude survey score as the DV and treatment/control, gender, and school as fixed factors. Treatment/control was nearly significant (p=.058), gender was nearly significant (p=.07), but school was not (p=.778). None of the interactions were significant. Does this mean that I should leave school out of further analysis?

2. Using the syntax pasted below, I ran another ANOVA, this time with attitude survey score (FullScaleRawScore) as the DV, treatment/control (PilotTeacher) as a fixed factor, and teacher (S08GroupNameRecoded) as a random factor nested within treatment/control. Can you verify that the way I wrote the syntax is correct? My interpretation of the results is that teacher nested within treatment/control is a significant predictor (p=.046) of attitude survey score but treatment/control is no longer important (p=.269) when teacher is accounted for. Is this interpretation correct?

3. If my interpretation in 2. is correct, how do I enter gender in the syntax?

Thanks for any help that you can provide. I really appreciate all the help that comes from this list.

UNIANOVA
FullScaleRawScore BY PilotTeacher S08GroupNameRecoded
/RANDOM = S08GroupNameRecoded
/METHOD = SSTYPE(1)
/DESIGN = PilotTeacher S08GroupNameRecoded WITHIN PilotTeacher

Tests of Between-Subjects Effects
Dependent Variable: FullScaleRawScore
| Source | | Type I Sum of Squares | df | Mean Square | F | Sig. |
| Intercept | Hypothesis | 837661.701 | 1 | 837661.701 | 3066.208 | .000 |
| | Error | 2590.063 | 9.481 | 273.191 | | |
| PilotTeacher | Hypothesis | 378.177 | 1 | 378.177 | 1.380 | .269 |
| | Error | 2580.716 | 9.415 | 274.098 | | |
| S08GroupNameRecoded(PilotTeacher) | Hypothesis | 2665.043 | 10 | 266.504 | 1.907 | .046 |
| | Error | 29215.078 | 209 | 139.785 | | |

____________________________________

Justin Meyer

Rowland Reading Foundation

phone: 866-370-7323 fax: 608-204-3846

____________________________________

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Marta García-Granero
Sent: Thursday, October 30, 2008 1:57 PM
To: [hidden email]
Subject: Re: Gender as Covariate in ANCOVA, Schools/Teachers as Dummy Variables, ANCOVA v. Regression?

Justin Meyer wrote:

Zdaniuk, Bozena-2

nested effects?

In reply to this post by Marta Garcia-Granero

I am sorry but I cannot wrap my mind around something: Why are cages nested within treatments if they were randomly assigned to treatments? Shouldn't the rats be nested within cages because they were not randomly assigned to cages?
What am I missing here?
Bozena

Bozena Zdaniuk, Ph.D.
University of Pittsburgh
UCSUR, 6th Fl.
121 University Place
Pittsburgh, PA 15260
Ph.: 412-624-5736
Fax: 412-624-4810
Email: [hidden email]

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Marta García-Granero
Sent: Thursday, October 30, 2008 4:04 PM
To: [hidden email]
Subject: Re: Gender as Covariate in ANCOVA, Schools/Teachers as Dummy Variables, ANCOVA v. Regression?

Zdaniuk, Bozena wrote:
> I have brought up (see below) a part of Marta's message I have a question about: I don't think you can designate factors as nested in the regular GLM ANOVA spss analysis, am I correct? I think one must use the Gneralized (as opposed to General) LM analysis. If I am wrong, please let me know how one can do nesting with the General LM.
> thanks.
>
You can use UNIANOVA, but not with the point&click interface, or you can
use MIXED:

* Sample dataset, 4 cages, with 4 rats each, are randomly assigned to 2
different ionizations 8negative 6 positive *.
DATA LIST LIST/ionization cage activity (3 F8).
BEGIN DATA
1 1 3
1 1 6
1 1 3
1 1 3
1 2 1
1 2 2
1 2 2
1 2 2
1 3 5
1 3 6
1 3 5
1 3 6
1 4 2
1 4 3
1 4 4
1 4 3
2 5 7
2 5 8
2 5 7
2 5 6
2 6 4
2 6 5
2 6 4
2 6 3
2 7 7
2 7 8
2 7 9
2 7 8
2 8 10
2 8 10
2 8 9
2 8 11
END DATA.

VALUE LABEL ionization 1'Negative' 2'Positive'.

UNIANOVA
activity BY ionization cage
/RANDOM = cage
/METHOD = SSTYPE(1)
/DESIGN = ionization cage WITHIN ionization .

MIXED
activity BY ionization cage
/FIXED = ionization | SSTYPE(1)
/METHOD = REML
/RANDOM cage(ionization) | COVTYPE(VC) .

HTH,
Marta

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

Marta Garcia-Granero

Re: nested effects?

Hi Bozena
> I am sorry but I cannot wrap my mind around something: Why are cages nested within treatments if they were randomly assigned to treatments? Shouldn't the rats be nested within cages because they were not randomly assigned to cages?
> What am I missing here?
>
Rats are assigned randomly to cages, and then cages are randomly
assigned to treatments. Cages form therefore a random factor, nested
inside treatments (cages 1 to 4 belong to treatment 1, and cages 5 to 8
to treatment 2). Rats don't form a factor, but within variation.

HTH,
Marta García-Granero

> * Sample dataset, 4 cages, with 4 rats each, are randomly assigned to 2
> different ionizations 8negative 6 positive *.
> DATA LIST LIST/ionization cage activity (3 F8).
> BEGIN DATA
> 1 1 3
> 1 1 6
> 1 1 3
> 1 1 3
> 1 2 1
> 1 2 2
> 1 2 2
> 1 2 2
> 1 3 5
> 1 3 6
> 1 3 5
> 1 3 6
> 1 4 2
> 1 4 3
> 1 4 4
> 1 4 3
> 2 5 7
> 2 5 8
> 2 5 7
> 2 5 6
> 2 6 4
> 2 6 5
> 2 6 4
> 2 6 3
> 2 7 7
> 2 7 8
> 2 7 9
> 2 7 8
> 2 8 10
> 2 8 10
> 2 8 9
> 2 8 11
> END DATA.
>
> VALUE LABEL ionization 1'Negative' 2'Positive'.
>
> UNIANOVA
> activity BY ionization cage
> /RANDOM = cage
> /METHOD = SSTYPE(1)
> /DESIGN = ionization cage WITHIN ionization .
>
> MIXED
> activity BY ionization cage
> /FIXED = ionization | SSTYPE(1)
> /METHOD = REML
> /RANDOM cage(ionization) | COVTYPE(VC) .
>
>
> HTH,
> Marta
>
> =====================
> 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
>
>

Zdaniuk, Bozena-2

Re: nested effects?

Ok, I get it (or so I hope:)). Now, the reason I asked was because I was trying to get around dyad problem.
I have caregiver-care recipient dyads that were randomly assigned to treatment or control and I measure them on five outcomes. Pearson correlation tells me that on two outcomes correlation between CG and CR score is significant. According to Kenny, I have no choice but to analyze at the level of dyad. But I was still hoping to analyze on the level of individuals to retain more power. So, I thought I would include a dyad id as random factor nested in the treatment and this way I can say that I remove the impact of dyad from the treatment effect. But it looks like I cannot do it because my individuals are not randomly assigned to dyads, right?
bozena

Bozena Zdaniuk, Ph.D.
University of Pittsburgh
UCSUR, 6th Fl.
121 University Place
Pittsburgh, PA 15260
Ph.: 412-624-5736
Fax: 412-624-4810
Email: [hidden email]

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Marta García-Granero
Sent: Thursday, November 06, 2008 2:11 PM
To: [hidden email]
Subject: Re: nested effects?

Hi Bozena
> I am sorry but I cannot wrap my mind around something: Why are cages nested within treatments if they were randomly assigned to treatments? Shouldn't the rats be nested within cages because they were not randomly assigned to cages?
> What am I missing here?
>
Rats are assigned randomly to cages, and then cages are randomly
assigned to treatments. Cages form therefore a random factor, nested
inside treatments (cages 1 to 4 belong to treatment 1, and cages 5 to 8
to treatment 2). Rats don't form a factor, but within variation.

HTH,
Marta García-Granero

--
For miscellaneous statistical stuff, visit:
http://gjyp.nl/marta/

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

Marta Garcia-Granero

Re: nested effects?

Zdaniuk, Bozena wrote:
> Ok, I get it (or so I hope:)). Now, the reason I asked was because I was trying to get around dyad problem.
> I have caregiver-care recipient dyads that were randomly assigned to treatment or control and I measure them on five outcomes. Pearson correlation tells me that on two outcomes correlation between CG and CR score is significant. According to Kenny, I have no choice but to analyze at the level of dyad. But I was still hoping to analyze on the level of individuals to retain more power. So, I thought I would include a dyad id as random factor nested in the treatment and this way I can say that I remove the impact of dyad from the treatment effect. But it looks like I cannot do it because my individuals are not randomly assigned to dyads, right?
>
Your design is not nested at all, but paired. Against your idea, if the
correlation between CG&CR score is significant, analyzing your data
using dyad as unit retains more power than treating your individuals
independently. Paired designs are very efficient since they remove
inter-pair heterogeneity. Just as a small proof: if you compute the
sample size you need to detect a standardized effect d=0.8 (Cohen's
threshold for large effects) as significant with a power of 80% in 2
independent samples design, you end up with 25 subjects per group. The
same calculation in a paired design will give you a figure close to 15
pairs. Therefore, a paired design is more efficient.

HTH,
Marta

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

Zdaniuk, Bozena-2

Re: nested effects?

Hm, (again, bear with me here) 15 pairs per group means 30 subjects per group. How is it more efficient than 25 subjects per group?

Bozena Zdaniuk, Ph.D.
University of Pittsburgh
UCSUR, 6th Fl.
121 University Place
Pittsburgh, PA 15260
Ph.: 412-624-5736
Fax: 412-624-4810
Email: [hidden email]

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Marta García-Granero
Sent: Thursday, November 06, 2008 3:00 PM
To: [hidden email]
Subject: Re: nested effects?

Zdaniuk, Bozena wrote:
> Ok, I get it (or so I hope:)). Now, the reason I asked was because I was trying to get around dyad problem.
> I have caregiver-care recipient dyads that were randomly assigned to treatment or control and I measure them on five outcomes. Pearson correlation tells me that on two outcomes correlation between CG and CR score is significant. According to Kenny, I have no choice but to analyze at the level of dyad. But I was still hoping to analyze on the level of individuals to retain more power. So, I thought I would include a dyad id as random factor nested in the treatment and this way I can say that I remove the impact of dyad from the treatment effect. But it looks like I cannot do it because my individuals are not randomly assigned to dyads, right?
>
Your design is not nested at all, but paired. Against your idea, if the
correlation between CG&CR score is significant, analyzing your data
using dyad as unit retains more power than treating your individuals
independently. Paired designs are very efficient since they remove
inter-pair heterogeneity. Just as a small proof: if you compute the
sample size you need to detect a standardized effect d=0.8 (Cohen's
threshold for large effects) as significant with a power of 80% in 2
independent samples design, you end up with 25 subjects per group. The
same calculation in a paired design will give you a figure close to 15
pairs. Therefore, a paired design is more efficient.

HTH,
Marta

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

Zdaniuk, Bozena-2

NORM help, please

In reply to this post by Justin Meyer-3

Hello, I am using Schafer's NORM software for multiple imputation and I ran into a problem I cannot figure out. I used it successfully about three years ago. This time, I thought I was following the same procedure but, for some weird reason, the program crashes on me every time I try to run data augmentation procedure (which produces multiple imputed files). I still run it successfully on those old data files I created three years ago so I am thinking there has to be something wrong with my current data file.
NORM only accepts ASCII (text) files. So, I create my file by saving it in SPSS using the 'save as tab delimited .dat file' option. Does anyone know if SPSS 15 does something to those files which makes them not 'pure' text files? I am running out of ideas here, as you can see...
Any help deeply appreciated.
Bozena

Bozena Zdaniuk, Ph.D.
University of Pittsburgh
UCSUR, 6th Fl.
121 University Place
Pittsburgh, PA 15260
Ph.: 412-624-5736
Fax: 412-624-4810
Email: [hidden email]

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

Re: NORM help, please

My first question would be whether NORM accepts only flat ASCII fixed
fields, or comma separated variable with single or double quotes around
strings that have spaces or punctuation embedded?

Try to read the old files with WordPad or whatever is the plain text
editor on your platform. If you click on various places on the tab bar
to create tabs and the data moves around on the screen, your old files
are tab-delimited. If there are commas (or pipes or semicolons) all
over the place you have a comma (or pipe or semicolon) separated file.
If neither of those work you most likely have a flat file. Within any of
the above see if any of your variables have spaces or punctuation in
some strings and if they do are there special delimiters?

Do the same with the new files.

If you still don't know what you have, try to read both old and new
files into excel.

Let us know if you are still stuck.

Art Kendall
Social Research Consultants

Zdaniuk, Bozena wrote:

> Hello, I am using Schafer's NORM software for multiple imputation and I ran into a problem I cannot figure out. I used it successfully about three years ago. This time, I thought I was following the same procedure but, for some weird reason, the program crashes on me every time I try to run data augmentation procedure (which produces multiple imputed files). I still run it successfully on those old data files I created three years ago so I am thinking there has to be something wrong with my current data file.
> NORM only accepts ASCII (text) files. So, I create my file by saving it in SPSS using the 'save as tab delimited .dat file' option. Does anyone know if SPSS 15 does something to those files which makes them not 'pure' text files? I am running out of ideas here, as you can see...
> Any help deeply appreciated.
> Bozena
>
> Bozena Zdaniuk, Ph.D.
> University of Pittsburgh
> UCSUR, 6th Fl.
> 121 University Place
> Pittsburgh, PA 15260
> Ph.: 412-624-5736
> Fax: 412-624-4810
> Email: [hidden email]
>
> =====================
> 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
Social Research Consultants

Zdaniuk, Bozena-2

Re: NORM help, please

In reply to this post by Zdaniuk, Bozena-2

Yes, I did that. That is not the issue. Any other ideas?
Bozena

-----Original Message-----
From: Mccoach, D. Betsy [mailto:[hidden email]]
Sent: Thursday, November 13, 2008 4:14 PM
To: Zdaniuk, Bozena
Subject: RE: NORM help, please

Make sure to unclick the checkbox that says "write variable names to
datafile", which is checked in SPSS by default.

Betsy

D. Betsy McCoach, Ph.D.
Associate Professor
Measurement, Evaluation, and Assessment Program
Educational Psychology Department
University of Connecticut
249 Glenbrook Road, Unit 2064
Storrs, CT 06269-2064
Phone: 860-486-0183
Fax: 860-486-0180
Email: [hidden email]

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Zdaniuk, Bozena
Sent: Thursday, November 13, 2008 4:04 PM
To: [hidden email]
Subject: NORM help, please

Hello, I am using Schafer's NORM software for multiple imputation and I
ran into a problem I cannot figure out. I used it successfully about
three years ago. This time, I thought I was following the same procedure
but, for some weird reason, the program crashes on me every time I try
to run data augmentation procedure (which produces multiple imputed
files). I still run it successfully on those old data files I created
three years ago so I am thinking there has to be something wrong with my
current data file.
NORM only accepts ASCII (text) files. So, I create my file by saving it
in SPSS using the 'save as tab delimited .dat file' option. Does anyone
know if SPSS 15 does something to those files which makes them not
'pure' text files? I am running out of ideas here, as you can see...
Any help deeply appreciated.
Bozena

Bozena Zdaniuk, Ph.D.
University of Pittsburgh
UCSUR, 6th Fl.
121 University Place
Pittsburgh, PA 15260
Ph.: 412-624-5736
Fax: 412-624-4810
Email: [hidden email]

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

Zdaniuk, Bozena-2

Re: NORM help, please

In reply to this post by Art Kendall

Thanks to all who tried to help. I just found out what was wrong and thought I would throw it out there in case someone runs into a similar problem.
The problem was the 12-digit id number. With other files, I could easily exclude the id from the model NORM runs but keep it in output files. For these data, maybe because the id was so long, even when I excluded it from the model, the program would crash every single time. Once I submitted data without the id number, NORM ran smoothly.
Bizarre..
Bozena

Bozena Zdaniuk, Ph.D.
University of Pittsburgh
UCSUR, 6th Fl.
121 University Place
Pittsburgh, PA 15260
Ph.: 412-624-5736
Fax: 412-624-4810
Email: [hidden email]

-----Original Message-----
From: Art Kendall [mailto:[hidden email]]
Sent: Thursday, November 13, 2008 4:51 PM
To: Zdaniuk, Bozena
Cc: [hidden email]
Subject: Re: NORM help, please

My first question would be whether NORM accepts only flat ASCII fixed
fields, or comma separated variable with single or double quotes around
strings that have spaces or punctuation embedded?

Try to read the old files with WordPad or whatever is the plain text
editor on your platform. If you click on various places on the tab bar
to create tabs and the data moves around on the screen, your old files
are tab-delimited. If there are commas (or pipes or semicolons) all
over the place you have a comma (or pipe or semicolon) separated file.
If neither of those work you most likely have a flat file. Within any of
the above see if any of your variables have spaces or punctuation in
some strings and if they do are there special delimiters?

Do the same with the new files.

If you still don't know what you have, try to read both old and new
files into excel.

Let us know if you are still stuck.

Art Kendall
Social Research Consultants

Zdaniuk, Bozena wrote:

Kreischer,Resha M

Does this bother anyone else? Problems Centering Output in 16 and 17

Hello,

I am trying to gauge if other SPSS users have noticed that it is
difficult to center output in SPSS 16 and 17 (I had no problems in
previous versions). I know you can export the output into Word, etc. but
it is even difficult to get the output centered there. I have notified
SPSS of the issue. Am I making a bigger deal out of this lack of basic
functionality than I should? How are others bypassing this issue?

Thanks,

Resha Kreischer-Anderson
Data Analyst

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

Re: Does this bother anyone else? Problems Centering Output in 16 and 17

When you specify centered output in Options, the centering applies only to
the printed output because only then does SPSS have the information as to
the width of the view. Turn on centering and look at your output in print
preview mode.

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Kreischer,Resha M
Sent: Tuesday, November 18, 2008 8:32 AM
To: [hidden email]
Subject: Does this bother anyone else? Problems Centering Output in 16 and
17

Hello,

I am trying to gauge if other SPSS users have noticed that it is
difficult to center output in SPSS 16 and 17 (I had no problems in
previous versions). I know you can export the output into Word, etc. but
it is even difficult to get the output centered there. I have notified
SPSS of the issue. Am I making a bigger deal out of this lack of basic
functionality than I should? How are others bypassing this issue?

Thanks,

Resha Kreischer-Anderson
Data Analyst

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

mpirritano

Re: Does this bother anyone else? Problems Centering Output in 16 and 17

In reply to this post by Kreischer,Resha M

Ahh, yes, I'd forgotten about this. I've experienced this as well and it
is very annoying but there is a simple workaround. For some reason when
you bring SPSS output into another program it comes in with a very odd
page size. I forget the exact size. Go to page setup and you'll see that
your page size has been set to some weird custom size. Just resize to
letter or a4, whatever you use, and then you'll be able to manipulate to
your hearts content.

Thanks
matt

Matthew Pirritano, Ph.D.
Research Analyst IV
Medical Services Initiative (MSI)
Orange County Health Care Agency
(714) 834-3566
-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Kreischer,Resha M
Sent: Tuesday, November 18, 2008 7:32 AM
To: [hidden email]
Subject: Does this bother anyone else? Problems Centering Output in 16
and 17

Hello,

I am trying to gauge if other SPSS users have noticed that it is
difficult to center output in SPSS 16 and 17 (I had no problems in
previous versions). I know you can export the output into Word, etc. but
it is even difficult to get the output centered there. I have notified
SPSS of the issue. Am I making a bigger deal out of this lack of basic
functionality than I should? How are others bypassing this issue?

Thanks,

Resha Kreischer-Anderson
Data Analyst

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