difference scores question
Locked
 
|
HI SPSSers,
I have a stats question, about difference scores. I need to look at convergence in attitudes between employees and their managers. The attitudes are measured with a 5-point Likert-type scale. I've done some research on the subject, and I understand there are a number of problems associated with merely taking the difference between them. I have identified three alternatives to calculating a difference score. They are: - use a residual score that is computed by taking the sub/mgr score and partialling out the effect of the other’s score - calculate a profile similarity index, or - use polynomial regression with response surface analysis. Could anyone tell me if any of the three is preferred, or if there is something else I should be looking at. As always, thank you for your time and help. Best, Lisa Lisa T. Stickney, Ph.D. Assistant Professor of Management Merrick School of Business University of Baltimore Ph: 410-837-6607 Fax: 410-837-5675 [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 |
|
|
Lisa,
One of the problems to consider with difference scores is that the measurement error in each variable is compounded when you compute their difference, especially if the true score variance in each measure is correlated (Cronbach did some work on this problem around the time that the original Cronbach and Furby critique of difference scores was published). To the extent that the variables you are comparing contain measurement error, you might want to consider using structural equation modeling to model the measurement errors as well as differences between levels of the latent variables. You may want to take the SEM approach if you residualize. Residualizing addresses some problems in difference scores (like correlations between level and change) but not the problems associated with measurement error. That being said, I have always liked polynomial trend surface analysis as it puts the differences between vaiables in the context of changesv in levels. Typically, a good visualization of the surface is needed ro communicate the results effectively. Best Steve Brand www.StatisticsDoc.com -----Original Message----- From: lts1 <[hidden email]> Date: Thu, 27 May 2010 09:13:42 To: <[hidden email]> Subject: difference scores question HI SPSSers, I have a stats question, about difference scores. I need to look at convergence in attitudes between employees and their managers. The attitudes are measured with a 5-point Likert-type scale. I've done some research on the subject, and I understand there are a number of problems associated with merely taking the difference between them. I have identified three alternatives to calculating a difference score. They are: - use a residual score that is computed by taking the sub/mgr score and partialling out the effect of the other’s score - calculate a profile similarity index, or - use polynomial regression with response surface analysis. Could anyone tell me if any of the three is preferred, or if there is something else I should be looking at. As always, thank you for your time and help. Best, Lisa Lisa T. Stickney, Ph.D. Assistant Professor of Management Merrick School of Business University of Baltimore Ph: 410-837-6607 Fax: 410-837-5675 [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 |
Re: difference scores question
|
|
In reply to this post by lts1
Lisa,
I wonder if your data analysis problem might be more complex although your sample size may not support the more complex analyses. If workers and managers rated the same 'thing', like work enviroment or productivity, it seems to me the data have a multilevel structure. The quantity you are interested in is the regression of the work groups employee's mean on the work groups manager rating (or vice versa). Possibly, another way to attack the problem is through a reciprocal social relations model (David Kenny is the key name). Gene Maguin -----Original Message----- From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of lts1 Sent: Thursday, May 27, 2010 9:14 AM To: [hidden email] Subject: difference scores question HI SPSSers, I have a stats question, about difference scores. I need to look at convergence in attitudes between employees and their managers. The attitudes are measured with a 5-point Likert-type scale. I've done some research on the subject, and I understand there are a number of problems associated with merely taking the difference between them. I have identified three alternatives to calculating a difference score. They are: - use a residual score that is computed by taking the sub/mgr score and partialling out the effect of the other's score - calculate a profile similarity index, or - use polynomial regression with response surface analysis. Could anyone tell me if any of the three is preferred, or if there is something else I should be looking at. As always, thank you for your time and help. Best, Lisa Lisa T. Stickney, Ph.D. Assistant Professor of Management Merrick School of Business University of Baltimore Ph: 410-837-6607 Fax: 410-837-5675 [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 |
|
In reply to this post by lts1
You're bound to get a professional statistical
answer from one of the listers, but have you considered Discriminant Function
Analysis to see if you can use the attitudes to predict which group respondents
belong to? Using single items may also be misleading: you need to look for
underlying dimensions and structure as well.
----- Original Message -----
|
Administrator
|
In reply to this post by lts1
Hi Lisa. I suspect that one of the problems you are referring to is the (purported) low reliability of difference scores. I started writing some notes on that a while ago, and this discussion prompted me to take another look at them. I did a little more work on them just now, and then uploaded them to my website. If you're interested, you can find them here:
https://sites.google.com/a/lakeheadu.ca/bweaver/Home/statistics/files/ReliabilityofDifferenceScores.pdf?attredirects=0&d=1 The main point is that the most commonly presented versions of the formula for the reliability of a difference make assumptions that often do not hold with real data; and with real data, the reliability may actually be a fair bit better than you might think. HTH. 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/). |
|
|
In reply to this post by statisticsdoc
Hi Lisa
Like Steve I think response surface analysis is a good way to go. However, I always have trouble interpreting the visualizations! (I also have no sense of direction by the way.) I prefer to understand what's going on by testing the significance of the linear difference and curvature along the line of perfect agreement, and the linear difference and curvature along the line of perfect disagreement. Garry Gelade Business Analytic Ltd -----Original Message----- From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Statisticsdoc Consulting Sent: 27 May 2010 14:57 To: [hidden email] Subject: Re: difference scores question Lisa, One of the problems to consider with difference scores is that the measurement error in each variable is compounded when you compute their difference, especially if the true score variance in each measure is correlated (Cronbach did some work on this problem around the time that the original Cronbach and Furby critique of difference scores was published). To the extent that the variables you are comparing contain measurement error, you might want to consider using structural equation modeling to model the measurement errors as well as differences between levels of the latent variables. You may want to take the SEM approach if you residualize. Residualizing addresses some problems in difference scores (like correlations between level and change) but not the problems associated with measurement error. That being said, I have always liked polynomial trend surface analysis as it puts the differences between vaiables in the context of changesv in levels. Typically, a good visualization of the surface is needed ro communicate the results effectively. Best Steve Brand www.StatisticsDoc.com -----Original Message----- From: lts1 <[hidden email]> Date: Thu, 27 May 2010 09:13:42 To: <[hidden email]> Subject: difference scores question HI SPSSers, I have a stats question, about difference scores. I need to look at convergence in attitudes between employees and their managers. The attitudes are measured with a 5-point Likert-type scale. I've done some research on the subject, and I understand there are a number of problems associated with merely taking the difference between them. I have identified three alternatives to calculating a difference score. They are: - use a residual score that is computed by taking the sub/mgr score and partialling out the effect of the other’s score - calculate a profile similarity index, or - use polynomial regression with response surface analysis. Could anyone tell me if any of the three is preferred, or if there is something else I should be looking at. As always, thank you for your time and help. Best, Lisa Lisa T. Stickney, Ph.D. Assistant Professor of Management Merrick School of Business University of Baltimore Ph: 410-837-6607 Fax: 410-837-5675 [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 ===================== 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 |
|
|
In reply to this post by lts1
Wouldn't the data also be confounded by the absolute level and dispersion of ratings between groups, making absolute difference that much harder to interpret. I think Gene alluded to that. Employees may be lower raters than managers, or more focused on primary needs (e.g., job security, benefits, salary) while managers might be more focused on career advancement or whatever.
I would think that each distribution first needs to be adjusted for this potential effect before differences can even be calculated... perhaps create standard scores for each attribute within each group, then calculate differences between groups and then assess convergence...? Bob Walker Surveys & Forecasts, LLC www.safllc.com -----Original Message----- From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of lts1 Sent: Thursday, May 27, 2010 9:14 AM To: [hidden email] Subject: difference scores question HI SPSSers, I have a stats question, about difference scores. I need to look at convergence in attitudes between employees and their managers. The attitudes are measured with a 5-point Likert-type scale. I've done some research on the subject, and I understand there are a number of problems associated with merely taking the difference between them. I have identified three alternatives to calculating a difference score. They are: - use a residual score that is computed by taking the sub/mgr score and partialling out the effect of the other’s score - calculate a profile similarity index, or - use polynomial regression with response surface analysis. Could anyone tell me if any of the three is preferred, or if there is something else I should be looking at. As always, thank you for your time and help. Best, Lisa Lisa T. Stickney, Ph.D. Assistant Professor of Management Merrick School of Business University of Baltimore Ph: 410-837-6607 Fax: 410-837-5675 [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 ======= Email scanned by PC Tools - No viruses or spyware found. (Email Guard: 7.0.0.18, Virus/Spyware Database: 6.15090) http://www.pctools.com/ ======= ======= Email scanned by PC Tools - No viruses or spyware found. (Email Guard: 7.0.0.18, Virus/Spyware Database: 6.15090) http://www.pctools.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 |
|
|
In reply to this post by Maguin, Eugene
Thanks Gene,
I hadn't considered a multi-level model, but I don't have groups of employees with a supervisor. The data are dyadic. Do you have any idea how small a "group" has to be to run a multi-level model? I would think a "group" of 1 is too small. As for a reciprocal social relations model, I'm not familiar with that, but I'll look into it. Thanks for the tip. Best, Lisa ----- Original Message ----- From: "Gene Maguin" <[hidden email]> To: <[hidden email]> Sent: Thursday, May 27, 2010 12:33 PM Subject: Re: difference scores question > Lisa, > > I wonder if your data analysis problem might be more complex although your > sample size may not support the more complex analyses. If workers and > managers rated the same 'thing', like work enviroment or productivity, it > seems to me the data have a multilevel structure. The quantity you are > interested in is the regression of the work groups employee's mean on the > work groups manager rating (or vice versa). Possibly, another way to > attack > the problem is through a reciprocal social relations model (David Kenny is > the key name). > > Gene Maguin > > > > -----Original Message----- > From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of > lts1 > Sent: Thursday, May 27, 2010 9:14 AM > To: [hidden email] > Subject: difference scores question > > HI SPSSers, > > I have a stats question, about difference scores. I need to look at > convergence in attitudes between employees and their managers. The > attitudes are measured with a 5-point Likert-type scale. I've done some > research on the subject, and I understand there are a number of problems > associated with merely taking the difference between them. I have > identified three alternatives to calculating a difference score. They are: > > - use a residual score that is computed by taking the sub/mgr score and > partialling out the effect of the other's score > - calculate a profile similarity index, or > - use polynomial regression with response surface analysis. > > Could anyone tell me if any of the three is preferred, or if there is > something else I should be looking at. > > As always, thank you for your time and help. > > Best, > Lisa > > > Lisa T. Stickney, Ph.D. > Assistant Professor of Management > Merrick School of Business > University of Baltimore > Ph: 410-837-6607 > Fax: 410-837-5675 > [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 > ===================== 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 |
|
|
In reply to this post by John F Hall
Hi John,
Nope. I hadn't considered this either. It may have
possibilities. I'll have to look into it. As for my measures being single items,
they are not. Supervisors and subordinates were given the same validated,
16-item scale to complete, and the reliabilities for both are good.
Thanks for your time.
Best,
Lisa
|
|
|
In reply to this post by Garry Gelade
Thanks Garry. How do you test the significance of the linear difference and
curvature along the line of perfect agreement, and the linear difference and curvature along the line of perfect disagreement? Best, Lisa ----- Original Message ----- From: "Garry Gelade" <[hidden email]> To: <[hidden email]> Sent: Thursday, May 27, 2010 4:13 PM Subject: Re: difference scores question > Hi Lisa > > Like Steve I think response surface analysis is a good way to go. > However, I always have trouble interpreting the visualizations! (I also > have no sense of direction by the way.) > > I prefer to understand what's going on by testing the significance of the > linear difference and curvature along the line of perfect agreement, and > the linear difference and curvature along the line of perfect > disagreement. > > Garry Gelade > Business Analytic Ltd > > -----Original Message----- > From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of > Statisticsdoc Consulting > Sent: 27 May 2010 14:57 > To: [hidden email] > Subject: Re: difference scores question > > Lisa, > One of the problems to consider with difference scores is that the > measurement error in each variable is compounded when you compute their > difference, especially if the true score variance in each measure is > correlated (Cronbach did some work on this problem around the time that > the original Cronbach and Furby critique of difference scores was > published). To the extent that the variables you are comparing contain > measurement error, you might want to consider using structural equation > modeling to model the measurement errors as well as differences between > levels of the latent variables. You may want to take the SEM approach if > you residualize. Residualizing addresses some problems in difference > scores (like correlations between level and change) but not the problems > associated with measurement error. > That being said, I have always liked polynomial trend surface analysis as > it puts the differences between vaiables in the context of changesv in > levels. Typically, a good visualization of the surface is needed ro > communicate the results effectively. > Best > Steve Brand > www.StatisticsDoc.com > > -----Original Message----- > From: lts1 <[hidden email]> > Date: Thu, 27 May 2010 09:13:42 > To: <[hidden email]> > Subject: difference scores question > > HI SPSSers, > > I have a stats question, about difference scores. I need to look at > convergence in attitudes between employees and their managers. The > attitudes are measured with a 5-point Likert-type scale. I've done some > research on the subject, and I understand there are a number of problems > associated with merely taking the difference between them. I have > identified three alternatives to calculating a difference score. They are: > > - use a residual score that is computed by taking the sub/mgr score and > partialling out the effect of the other’s score > - calculate a profile similarity index, or > - use polynomial regression with response surface analysis. > > Could anyone tell me if any of the three is preferred, or if there is > something else I should be looking at. > > As always, thank you for your time and help. > > Best, > Lisa > > > Lisa T. Stickney, Ph.D. > Assistant Professor of Management > Merrick School of Business > University of Baltimore > Ph: 410-837-6607 > Fax: 410-837-5675 > [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 > > ===================== > 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 |
|
|
In reply to this post by lts1
Lisa
Gene's suggestion is a good one. Your instance more closely resembles the use of HLM to model dyadic data (such as couples or siblings) rather than groups. Think of the difference between supervisor and employee as a within dyad slope. HTH Steve Brand www.StatisticsDoc.com -----Original Message----- From: "Lisa T. Stickney" <[hidden email]> Date: Fri, 28 May 2010 18:02:12 To: <[hidden email]> Subject: Re: difference scores question Thanks Gene, I hadn't considered a multi-level model, but I don't have groups of employees with a supervisor. The data are dyadic. Do you have any idea how small a "group" has to be to run a multi-level model? I would think a "group" of 1 is too small. As for a reciprocal social relations model, I'm not familiar with that, but I'll look into it. Thanks for the tip. Best, Lisa ----- Original Message ----- From: "Gene Maguin" <[hidden email]> To: <[hidden email]> Sent: Thursday, May 27, 2010 12:33 PM Subject: Re: difference scores question > Lisa, > > I wonder if your data analysis problem might be more complex although your > sample size may not support the more complex analyses. If workers and > managers rated the same 'thing', like work enviroment or productivity, it > seems to me the data have a multilevel structure. The quantity you are > interested in is the regression of the work groups employee's mean on the > work groups manager rating (or vice versa). Possibly, another way to > attack > the problem is through a reciprocal social relations model (David Kenny is > the key name). > > Gene Maguin > > > > -----Original Message----- > From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of > lts1 > Sent: Thursday, May 27, 2010 9:14 AM > To: [hidden email] > Subject: difference scores question > > HI SPSSers, > > I have a stats question, about difference scores. I need to look at > convergence in attitudes between employees and their managers. The > attitudes are measured with a 5-point Likert-type scale. I've done some > research on the subject, and I understand there are a number of problems > associated with merely taking the difference between them. I have > identified three alternatives to calculating a difference score. They are: > > - use a residual score that is computed by taking the sub/mgr score and > partialling out the effect of the other's score > - calculate a profile similarity index, or > - use polynomial regression with response surface analysis. > > Could anyone tell me if any of the three is preferred, or if there is > something else I should be looking at. > > As always, thank you for your time and help. > > Best, > Lisa > > > Lisa T. Stickney, Ph.D. > Assistant Professor of Management > Merrick School of Business > University of Baltimore > Ph: 410-837-6607 > Fax: 410-837-5675 > [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 > ===================== 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 |
|
|
Lisa,
This post is intended to help you fit the model employing the MIXED procedure. It is not intended to address all of your concerns. It is possible that there is a better approach.
Employees and managers are nested in dyads. Construct your data set in long format as follows:
Person_ID Dyad_ID Position Y
1 1 1 5
2 1 2 3
3 2 1 2
4 2 2 1
5 3 1 3
6 3 2 4
.
.
.
N
Person_ID = Person Indictator Variable
Dyad_ID = Dyad Indicator Variable
Position = Position Indicator Variable (1=Employee, 2=Manager)
Y = Dependent Variable
To test if there is a significant mean difference between positions, use the following code:
MIXED Y BY Position
/FIXED=Position | SSTYPE(3) /METHOD=REML /PRINT=SOLUTION /RANDOM=INTERCEPT | SUBJECT(Dyad_ID) COVTYPE(VC). The random intercept term accounts for covariation of persons in the same dyad. The "Estimates of Fixed Effects" table in the output file will provide you with the mean difference test.
Ryan On Fri, May 28, 2010 at 6:13 PM, Statisticsdoc Consulting <[hidden email]> wrote: Lisa |
|
|
Thanks everyone. I appreciate your help. You've
given me several good ideas and much to ponder. As always, you're the
best.
Lisa
|
|
|
In reply to this post by lts1
There is an extensive collection of user-contributed syntax and macros for dyadic analysis available from SPSS Developer Central (www.spss.com/devcentral)
HTH, Jon Peck (From Florence) ----------------- Sent from my BlackBerry Handheld. ----- Original Message ----- From: "Lisa T. Stickney" [[hidden email]] Sent: 05/28/2010 06:02 PM AST To: [hidden email] Subject: Re: [SPSSX-L] difference scores question Thanks Gene, I hadn't considered a multi-level model, but I don't have groups of employees with a supervisor. The data are dyadic. Do you have any idea how small a "group" has to be to run a multi-level model? I would think a "group" of 1 is too small. As for a reciprocal social relations model, I'm not familiar with that, but I'll look into it. Thanks for the tip. Best, Lisa ----- Original Message ----- From: "Gene Maguin" <[hidden email]> To: <[hidden email]> Sent: Thursday, May 27, 2010 12:33 PM Subject: Re: difference scores question > Lisa, > > I wonder if your data analysis problem might be more complex although your > sample size may not support the more complex analyses. If workers and > managers rated the same 'thing', like work enviroment or productivity, it > seems to me the data have a multilevel structure. The quantity you are > interested in is the regression of the work groups employee's mean on the > work groups manager rating (or vice versa). Possibly, another way to > attack > the problem is through a reciprocal social relations model (David Kenny is > the key name). > > Gene Maguin > > > > -----Original Message----- > From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of > lts1 > Sent: Thursday, May 27, 2010 9:14 AM > To: [hidden email] > Subject: difference scores question > > HI SPSSers, > > I have a stats question, about difference scores. I need to look at > convergence in attitudes between employees and their managers. The > attitudes are measured with a 5-point Likert-type scale. I've done some > research on the subject, and I understand there are a number of problems > associated with merely taking the difference between them. I have > identified three alternatives to calculating a difference score. They are: > > - use a residual score that is computed by taking the sub/mgr score and > partialling out the effect of the other's score > - calculate a profile similarity index, or > - use polynomial regression with response surface analysis. > > Could anyone tell me if any of the three is preferred, or if there is > something else I should be looking at. > > As always, thank you for your time and help. > > Best, > Lisa > > > Lisa T. Stickney, Ph.D. > Assistant Professor of Management > Merrick School of Business > University of Baltimore > Ph: 410-837-6607 > Fax: 410-837-5675 > [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 > ===================== 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 |
«
Return to SPSSX Discussion
|
1 view|%1 views
| Free forum by Nabble | Edit this page |