Fwd: Latent Variable Analysis: CFA, EFA, and IRT?
Locked
Fwd: Latent Variable Analysis: CFA, EFA, and IRT?
 
|
---------- Forwarded message ---------- From: Deena Isom <[hidden email]> Date: Mon, Jan 20, 2014 at 4:38 PM Subject: Latent Variable Analysis: CFA, EFA, and IRT? To: [hidden email] Hello List, I'm trying to determine which analytical strategy is best for a current project, and if any are adequately conducted in SPSS or if another stats package is needed.
I'm trying to assess if two latent constructs are truly distinct, when in most literature items assessing each they are measured together in one scale..... I believe this is best done with exploratory factor analysis to see how different questions load of different factors, but if I believe they are different, would confirmatory factor analysis better serve these purposes - force their loadings? Also, can either of these really be fully conducted in SPSS and provide information about fit and so forth? And finally, could item response theory help distinguish which questions measure which concepts? Honestly, I'm still confused by IRT.
Thanks in advance for your insight. Hope you had a wonderful holiday! -- ~Deena -- ~Deena |
Re: Latent Variable Analysis: CFA, EFA, and IRT?
|
|
1.
You can use either EFA or CFA for this study. CFA might be better because you have already had a structure in mind. Also, EFA involves much more subjectivity
than does CFA. 2.
SPSS Statistics for EFA. SPSS AMOS for EFA (under Specification Search) and CFA. Hongwei Yang, University of Kentucky From: SPSSX(r) Discussion [mailto:[hidden email]]
On Behalf Of Deena Isom ---------- Forwarded message ---------- Hello List, I'm trying to determine which analytical strategy is best for a current project, and if any are adequately conducted in SPSS or if another stats package is needed.
I'm trying to assess if two latent constructs are truly distinct, when in most literature items assessing each they are measured together in one scale..... I believe this is best done with exploratory factor analysis to see how different
questions load of different factors, but if I believe they are different, would confirmatory factor analysis better serve these purposes - force their loadings? Also, can either of these really be fully conducted in SPSS and provide information about fit
and so forth? And finally, could item response theory help distinguish which questions measure which concepts? Honestly, I'm still confused by IRT.
Thanks in advance for your insight. Hope you had a wonderful holiday!
|
Re: Fwd: Latent Variable Analysis: CFA, EFA, and IRT?
|
In reply to this post by Deena Isom
Please
give more details.
Do you mean that the instrument has two scales?. I.e., some items go into one score and other items go into another score? Or do you mean that the he instrument produces one score and you think it should produce 2 scores. What constructs is the instrument designed to measure? What constructs do you think it should measure? Has the instrument be psychometrically validated? How did the authors of the instrument validate the score? How many items are there? What is the response scale? What are the items and how do you think they should be grouped? How many cases do you have? Art Kendall Social Research ConsultantsOn 1/20/2014 4:46 PM, Deena Isom [via SPSSX Discussion] wrote:
Art Kendall
Social Research Consultants |
Re: Latent Variable Analysis: CFA, EFA, and IRT?
|
|
In reply to this post by Deena Isom
This is a problem that can only be done on a SEM program such as Amos. As I understand your message, you are asking is whether a two factor model fits better
than a one factor model. And, you propose that factor 1 is composed of items A, B, D and factor 2 is composed of items C, E, F. (You can force a two factor solution but you can not control item-to-factor assignment.) The one factor model should have a df of
9 and the two factor model should have a df of 8. I think, but I may be wrong, that if you constrain the factor variances in the two factor model to be equal, you will have the correct nested model test.
Gene Maguin From: SPSSX(r) Discussion [mailto:[hidden email]]
On Behalf Of Deena Isom ---------- Forwarded message ---------- Hello List, I'm trying to determine which analytical strategy is best for a current project, and if any are adequately conducted in SPSS or if another stats package is needed.
I'm trying to assess if two latent constructs are truly distinct, when in most literature items assessing each they are measured together in one scale..... I believe this is best done with exploratory factor analysis to see how different
questions load of different factors, but if I believe they are different, would confirmatory factor analysis better serve these purposes - force their loadings? Also, can either of these really be fully conducted in SPSS and provide information about fit
and so forth? And finally, could item response theory help distinguish which questions measure which concepts? Honestly, I'm still confused by IRT.
Thanks in advance for your insight. Hope you had a wonderful holiday!
|
Re: Fwd: Latent Variable Analysis: CFA, EFA, and IRT?
|
|
In reply to this post by Deena Isom
If you have a strong background, and good reasoning then CFA is the only way to be sure that your model fits. However, that requires either expensive extra software or using ad hoc software. Otherwise, if you have a good may many more cases than variables (10x?), a PAF with varimax rotation will give you a good shot at scales that have discriminative validity. Be sure to keep only items that load cleanly if you retain more than 1 factor. It is possible that you will need to generate a larger items pool if you want to have items that measure more constructs. Parallel analysis will help you determine the number of factors to retain. Any thing more specific would depend on knowing the actual items, the eigenvalues, etc. Art Kendall Social Research ConsultantsOn 1/20/2014 4:46 PM, Deena Isom [via SPSSX Discussion] wrote:
Art Kendall
Social Research Consultants |
Re: Fwd: Latent Variable Analysis: CFA, EFA, and IRT?
Administrator
|
"However, that requires either expensive extra software or using ad hoc software. "
Not necessarily true: AFAIK there are a few FREE SEM solutions MX, LISREL has a free student version limited number of variables, there is an R package called lavaan (sp). Probably others but I have not investigated. OP take the time to do a little web search and you will likely locate them.
Please reply to the list and not to my personal email.
Those desiring my consulting or training services please feel free to email me. --- "Nolite dare sanctum canibus neque mittatis margaritas vestras ante porcos ne forte conculcent eas pedibus suis." Cum es damnatorum possederunt porcos iens ut salire off sanguinum cliff in abyssum?" |
Re: Fwd: Latent Variable Analysis: CFA, EFA, and IRT?
|
|
I suspect AMOS is sufficient for the problem here.
Hongwei Yang, University of Kentucky -----Original Message----- From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of David Marso Sent: Tuesday, January 21, 2014 3:55 PM To: [hidden email] Subject: Re: Fwd: Latent Variable Analysis: CFA, EFA, and IRT? "However, that requires either expensive extra software or using ad hoc software. " Not necessarily true: AFAIK there are a few FREE SEM solutions MX, LISREL has a free student version limited number of variables, there is an R package called lavaan (sp). Probably others but I have not investigated. OP take the time to do a little web search and you will likely locate them. Art Kendall wrote > If you have a strong background, and good reasoning then CFA is > the only way to be sure that your model fits. > However, that requires either expensive extra software or using > ad hoc software. > > Otherwise, if you have a good may many more cases than variables > (10x?), a PAF with varimax rotation will give you a good shot at > scales that have discriminative validity. Be sure to keep only > items that load cleanly if you retain more than 1 factor. > > It is possible that you will need to generate a larger items > pool if you want to have items that measure more constructs. > > Parallel analysis will help you determine the number of factors > to retain. > > > > Any thing more specific would depend on knowing the actual > items, the eigenvalues, etc. > > Art Kendall > Social Research Consultants > On 1/20/2014 4:46 PM, Deena Isom [via SPSSX Discussion] wrote: > > > > > ---------- Forwarded message ---------- > From: Deena Isom < [hidden > email] > > Date: Mon, Jan 20, 2014 at 4:38 PM > Subject: Latent Variable Analysis: CFA, EFA, and IRT? > To: [hidden email] > > > > Hello List, > > > I'm trying to determine which analytical strategy is > best for a current project, and if any are adequately > conducted in SPSS or if another stats package is needed. > > I'm trying to assess if two latent constructs are truly > distinct, when in most literature items assessing > each they are measured together in one scale..... I > believe this is best done with exploratory factor > analysis to see how different questions load of different > factors, but if I believe they are different, would > confirmatory factor analysis better serve these purposes - > force their loadings? Also, can either of these really be > fully conducted in SPSS and provide information about fit > and so forth? And finally, could item response theory > help distinguish which questions measure which concepts? > Honestly, I'm still confused by IRT. > > Thanks in advance for your insight. Hope you had a > wonderful holiday! > > > > -- > ~Deena > > > > > > > -- > ~Deena > > > > > > If you reply to this email, your > message will be added to the discussion below: > > http://spssx-discussion.1045642.n5.nabble.com/Fwd-Latent-Variable-Anal > ysis-CFA-EFA-and-IRT-tp5724030.html > > > To start a new topic under SPSSX Discussion, email > > ml-node+s1045642n1068821h68@.nabble > > To unsubscribe from SPSSX Discussion, click > here . > NAML ----- Please reply to the list and not to my personal email. Those desiring my consulting or training services please feel free to email me. --- "Nolite dare sanctum canibus neque mittatis margaritas vestras ante porcos ne forte conculcent eas pedibus suis." Cum es damnatorum possederunt porcos iens ut salire off sanguinum cliff in abyssum?" -- View this message in context: http://spssx-discussion.1045642.n5.nabble.com/Fwd-Latent-Variable-Analysis-CFA-EFA-and-IRT-tp5724030p5724050.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: Fwd: Latent Variable Analysis: CFA, EFA, and IRT?
|
|
In reply to this post by Art Kendall
Thanks everyone! That was helpful. On Tue, Jan 21, 2014 at 1:53 PM, Art Kendall <[hidden email]> wrote:
-- ~Deena |
«
Return to SPSSX Discussion
|
1 view|%1 views
| Free forum by Nabble | Edit this page |