Indicator Reliability, Redundancy, Bootstrapping
-
- PLS Expert User
- Posts: 31
- Joined: Mon Jun 14, 2010 7:35 pm
- Real name and title:
Indicator Reliability, Redundancy, Bootstrapping
Dear all,
at the moment I´m writing my diploma thesis in marketing.
I have a model with formative exogenous indicators and reflektive endogenous indicators and have started to examine my model with PLS. Now I have some questions about my output:
1. Where do I have to look for the factor loadings of the reflektive indicators? outer loadings or weights or loadings? The same questions for the formative indicators.
2. If I look at the redundancies I see, that they are all negative. I think that this is bad becausé it reflects the Stone Geisser Criteria and this needs to be above 0. What do I have to do now to get better reduncancies?
3. How do I do the Bootstrapping? I have a sample of 274, what are the settings for cases and samples? When I have done the Bootstrapping, do I have to look for reduncany there? I have some difficulties in understanding the output and where to look for the right output. Where do I see the T-values and significance of the indicators (question 1)?
I hope you understand all of my problems :)
I would be very grateful if somebody could help me.
Thank you very much.
Beste regards,
Larissa Lindemann
at the moment I´m writing my diploma thesis in marketing.
I have a model with formative exogenous indicators and reflektive endogenous indicators and have started to examine my model with PLS. Now I have some questions about my output:
1. Where do I have to look for the factor loadings of the reflektive indicators? outer loadings or weights or loadings? The same questions for the formative indicators.
2. If I look at the redundancies I see, that they are all negative. I think that this is bad becausé it reflects the Stone Geisser Criteria and this needs to be above 0. What do I have to do now to get better reduncancies?
3. How do I do the Bootstrapping? I have a sample of 274, what are the settings for cases and samples? When I have done the Bootstrapping, do I have to look for reduncany there? I have some difficulties in understanding the output and where to look for the right output. Where do I see the T-values and significance of the indicators (question 1)?
I hope you understand all of my problems :)
I would be very grateful if somebody could help me.
Thank you very much.
Beste regards,
Larissa Lindemann
-
- PLS Expert User
- Posts: 248
- Joined: Sat Jul 25, 2009 1:34 pm
- Real name and title:
Dear Larissa,
All beginnings are difficult.
Some hints:
ad 1) Firstly, you can find the loadings for reflektive indicators and the weights for formative indicators in your graphic, after you have run the normal ‘PLS Algorithms’ calculation. Secondly, you will also find the necessary information in your ‘Default Report’ after you have run the ‘PLS Algorithms’ under ‘Outer Loadings’ and ‘Outer Weights’ for your indicators. To get the PLS “Default Report” you have to click the button with the pencil on a sheet (right beside the calculation button).
ad 2) There is a small bug in the blindfolding procedure at the actual release, as reported elsewhere in this forum (s. viewtopic.php?t=944). Furthermore, you have to look at the right place at the output: You find the right output in SmartPLS at the “Construct Crossvalidated Redundancy” output. There you have to look at the block with the “Total “ sums. Perhaps you results are changing. You can find some more information about this important criteria in the following working paper (only in german): http://www.ibl-unihh.de/ap21_Stand_Juni2010.pdf
ad 3) For the Bootstrapping settings you have to use the size of your sample (274) for ‘cases’ and 1000 or so for ‘samples’. After you have run the Bootstraping procedure you find in default report the necessary t-values for the weights and for the path coefficients.
I hope this helps!
Best regards,
Christian
All beginnings are difficult.
Some hints:
ad 1) Firstly, you can find the loadings for reflektive indicators and the weights for formative indicators in your graphic, after you have run the normal ‘PLS Algorithms’ calculation. Secondly, you will also find the necessary information in your ‘Default Report’ after you have run the ‘PLS Algorithms’ under ‘Outer Loadings’ and ‘Outer Weights’ for your indicators. To get the PLS “Default Report” you have to click the button with the pencil on a sheet (right beside the calculation button).
ad 2) There is a small bug in the blindfolding procedure at the actual release, as reported elsewhere in this forum (s. viewtopic.php?t=944). Furthermore, you have to look at the right place at the output: You find the right output in SmartPLS at the “Construct Crossvalidated Redundancy” output. There you have to look at the block with the “Total “ sums. Perhaps you results are changing. You can find some more information about this important criteria in the following working paper (only in german): http://www.ibl-unihh.de/ap21_Stand_Juni2010.pdf
ad 3) For the Bootstrapping settings you have to use the size of your sample (274) for ‘cases’ and 1000 or so for ‘samples’. After you have run the Bootstraping procedure you find in default report the necessary t-values for the weights and for the path coefficients.
I hope this helps!
Best regards,
Christian
Last edited by christian.nitzl on Mon Sep 06, 2010 3:16 pm, edited 1 time in total.
-
- PLS Expert User
- Posts: 31
- Joined: Mon Jun 14, 2010 7:35 pm
- Real name and title:
Dear Cristian/all,
I am new user smartPLS and currently on the phase of analyzing my theoretical model using smartPLS for my final project. If you do not mind, I would like want several question concerning using and interpretation output (I use the paper from Liang et al., (2007), Assimilation of Enterprise Systems: The Effect of Institutional Pressure and the Mediating Role of Top Management, MIS Quarterly 31(1), pp. 59-87.) :
1. Relate to Larissa 1st question, on Liang et al., (2007) appendix C, they called it as item loadings and Cross loading, while other paper called it as factor loading. My question is: Is it the “cross loading” output below the quality criteria that they mean?
2. Concerning the Bootstrapping, which T-value should we use for assuring the confirm or not confirm relation among two variables/constructs?
3. Regarding the common method bias, in which report output I can find the “substantive factor loading (R1), R12, “Method factor loading(R22)”, and R22 (Appendix D on Liang et al (2007) paper)?
Thank you in advance and I am very grateful for your answer.
Salam,
Tettri Nuraini
I am new user smartPLS and currently on the phase of analyzing my theoretical model using smartPLS for my final project. If you do not mind, I would like want several question concerning using and interpretation output (I use the paper from Liang et al., (2007), Assimilation of Enterprise Systems: The Effect of Institutional Pressure and the Mediating Role of Top Management, MIS Quarterly 31(1), pp. 59-87.) :
1. Relate to Larissa 1st question, on Liang et al., (2007) appendix C, they called it as item loadings and Cross loading, while other paper called it as factor loading. My question is: Is it the “cross loading” output below the quality criteria that they mean?
2. Concerning the Bootstrapping, which T-value should we use for assuring the confirm or not confirm relation among two variables/constructs?
3. Regarding the common method bias, in which report output I can find the “substantive factor loading (R1), R12, “Method factor loading(R22)”, and R22 (Appendix D on Liang et al (2007) paper)?
Thank you in advance and I am very grateful for your answer.
Salam,
Tettri Nuraini
Hi Tettri,
a short answer to your questions:
1. Correct, that is the table Liang et al. used in their work (cross-loadings under quality criteria in SmartPLS)
2. Use the T-Value reported by SmartPLS (t-Values of path coefficients in the default report after the boostrap procedure) -> use individual sign changes as an option (see Teenenhaus, 2005 for further details on that issue)
3. You wont find this report in Smart PLS. But you can create it by yourself. Just proceed (transform your model) as Liang et al. explain in Appendix E. Then run the model and use the path coefficients as explained earlier in this thread. If you multiply the loadings (path coefficients) then you get the R12 and the R22.
Hope this helps overall. Just shoot for further questions.
KR
Mat
a short answer to your questions:
1. Correct, that is the table Liang et al. used in their work (cross-loadings under quality criteria in SmartPLS)
2. Use the T-Value reported by SmartPLS (t-Values of path coefficients in the default report after the boostrap procedure) -> use individual sign changes as an option (see Teenenhaus, 2005 for further details on that issue)
3. You wont find this report in Smart PLS. But you can create it by yourself. Just proceed (transform your model) as Liang et al. explain in Appendix E. Then run the model and use the path coefficients as explained earlier in this thread. If you multiply the loadings (path coefficients) then you get the R12 and the R22.
Hope this helps overall. Just shoot for further questions.
KR
Mat
Hi Christian, hi all
could someone please help me. Reading the topic I thought it would be exactly what I need.
Actually it wasn't, so my question is:
In which output I could find the "indicator reliability". The construct reliability is in the PLS-Overview-Output. But where do I find the reliability for the indicators?
Many thx
Angel
could someone please help me. Reading the topic I thought it would be exactly what I need.
Actually it wasn't, so my question is:
In which output I could find the "indicator reliability". The construct reliability is in the PLS-Overview-Output. But where do I find the reliability for the indicators?
Many thx
Angel
-
- PLS Expert User
- Posts: 248
- Joined: Sat Jul 25, 2009 1:34 pm
- Real name and title:
Hi Christian,
thx for your immediate answer. I was just wondering, because in several paper I could see both factor loadings and indicator reliability. So I assumed it have to be any difference between both.
e.g. unfortn. only in German:
Clement, M./Völckner, F./Granström, N./Dyk, T. van (2008): Messung der Marken-stärke von Künstlermarken. Eine empirische Untersuchung am Beispiel von Popmusi-kern. In: Marketing Zeitschrift für Forschung und Praxis, 30 (2), S. 93-108.
Do you have any idea? Isn't there a similar Output in smartpls?
Regards and thx in advance
Angel
thx for your immediate answer. I was just wondering, because in several paper I could see both factor loadings and indicator reliability. So I assumed it have to be any difference between both.
e.g. unfortn. only in German:
Clement, M./Völckner, F./Granström, N./Dyk, T. van (2008): Messung der Marken-stärke von Künstlermarken. Eine empirische Untersuchung am Beispiel von Popmusi-kern. In: Marketing Zeitschrift für Forschung und Praxis, 30 (2), S. 93-108.
Do you have any idea? Isn't there a similar Output in smartpls?
Regards and thx in advance
Angel
-
- PLS Expert User
- Posts: 248
- Joined: Sat Jul 25, 2009 1:34 pm
- Real name and title:
Hey Angel,
The problem is that have not the possibility to get the cited paper as an e-paper.
If you can post here the table as a picture perhaps I could help you. But I believe that some authors square the loadings for an easier interpretation of the indicator reliability. It is said that the squared loadings should be higher than 0.5
Best regards,
Christian
The problem is that have not the possibility to get the cited paper as an e-paper.
If you can post here the table as a picture perhaps I could help you. But I believe that some authors square the loadings for an easier interpretation of the indicator reliability. It is said that the squared loadings should be higher than 0.5
Best regards,
Christian
Hi Christian,
see below one table from above paper:
Faktorladung Item-to-Total- Indikatorreliabilität
Korrelation
0,77 0,60 0,53
0,87 0,73 0,74
0,85 0,70 0,75
0,79 0,64 0,65
So unfortunately it doesn't seem to be the solution to square the loadings.
Maybe you have any other idea?
Thx so much for your help.
Regards
Angel
see below one table from above paper:
Faktorladung Item-to-Total- Indikatorreliabilität
Korrelation
0,77 0,60 0,53
0,87 0,73 0,74
0,85 0,70 0,75
0,79 0,64 0,65
So unfortunately it doesn't seem to be the solution to square the loadings.
Maybe you have any other idea?
Thx so much for your help.
Regards
Angel
-
- PLS Expert User
- Posts: 248
- Joined: Sat Jul 25, 2009 1:34 pm
- Real name and title:
Hey Angel,
It seems that Clement et al. use an variance based approach (e.g. LISREL, AMOS, MPlus) for their study. Therefore that should not bother you. In covariance based approach it is possible to use some more criteria for checking measurements and model because the assumption of normal distribution. In an variance based approach as PLS we can only use a part of this criteria e.g. Loadings, AVE, Composite Reliability.
Do you know my working paper (only German)? In the working paper you can find all information to get happy with PLS :-)
Here comes the link: http://www.ibl-unihh.de/ap21_Stand_Juni2010.pdf
On pp. 23-39 you find all important criteria for evaluate the measurements and the structure model in a PLS analysis.
Best regards,
Christian
It seems that Clement et al. use an variance based approach (e.g. LISREL, AMOS, MPlus) for their study. Therefore that should not bother you. In covariance based approach it is possible to use some more criteria for checking measurements and model because the assumption of normal distribution. In an variance based approach as PLS we can only use a part of this criteria e.g. Loadings, AVE, Composite Reliability.
Do you know my working paper (only German)? In the working paper you can find all information to get happy with PLS :-)
Here comes the link: http://www.ibl-unihh.de/ap21_Stand_Juni2010.pdf
On pp. 23-39 you find all important criteria for evaluate the measurements and the structure model in a PLS analysis.
Best regards,
Christian
-
- PLS Expert User
- Posts: 248
- Joined: Sat Jul 25, 2009 1:34 pm
- Real name and title:
Hey,
I’m so sorry but I have the working paper only in German. May I recommend you three recent published article in English which are very helpful for the understanding of PLS:
J. F. Hair; C. M. Ringle; M. Sarstedt (2011): PLS-SEM. Indeed a Silver Bullet, Journal of Marketing Theory & Practice, Vol. 19 (2), S. 139-151.
E. E. Rigdon; C. M. Ringle; M. Sarstedt (2010): Structural Modeling of Heterogeneous Data with Partial Least Squares, Review of Marketing Research, Vol. 7, S. 255 - 296.
Henseler, J./ Ringle, C.M./ Sinkovics, R.R.: The use of partial least squares path modeling in international marketing, in: Sinkovics, R.R./ Ghauri, P.N. (eds.), Advances in International Marketing (AIM), Volume 20, Bingley 2009, pp. 277-320 (http://php.portals.mbs.ac.uk/Portals/49 ... cs-PLS.pdf).
I hope that helps!
Christian
I’m so sorry but I have the working paper only in German. May I recommend you three recent published article in English which are very helpful for the understanding of PLS:
J. F. Hair; C. M. Ringle; M. Sarstedt (2011): PLS-SEM. Indeed a Silver Bullet, Journal of Marketing Theory & Practice, Vol. 19 (2), S. 139-151.
E. E. Rigdon; C. M. Ringle; M. Sarstedt (2010): Structural Modeling of Heterogeneous Data with Partial Least Squares, Review of Marketing Research, Vol. 7, S. 255 - 296.
Henseler, J./ Ringle, C.M./ Sinkovics, R.R.: The use of partial least squares path modeling in international marketing, in: Sinkovics, R.R./ Ghauri, P.N. (eds.), Advances in International Marketing (AIM), Volume 20, Bingley 2009, pp. 277-320 (http://php.portals.mbs.ac.uk/Portals/49 ... cs-PLS.pdf).
I hope that helps!
Christian