Hi Konrad. First understand that it is IBM AMOS.24, Not SPSS AMOS. (till 17 it was SPSS AMOS).

I watched your model, It can be tested in AMOS separately or at time. I can't understand ur question clearly what u mean. Let me know about your research objectives and statistical analysis what you basically want to perform so.? 

Regards

Dear Konrad,

The similar problem has been solved and explained very clearly in the paper given below:Amoako-Gyampah, K. and Acquaah, M. (2008) Manufacturing strategy, competitive strategy and firm performance: An empirical study in a developing economy environment. International Journal of Production Economics, 111(2): 575-592.

The paper explains the concept of direct, indirect and total effect.

Here the Indirect effect of A on B is calculated as: Direct effect of (A on M1) * Direct effect of (M1 on B).

Total Effect is = Direct + inderect.

I recommend to go through this paper for more clarity.

Best wishes.

Thanks for all responses!

Answar, thank you for corrected me and giving me a proper name of AMOS software.

Manjeet, thank you very much for references to Amoako-Gyampah and Acquaah, (2008) interesting paper. However, Amoako-Gyampah and Acquaah (2008) simply multiply specific direct effect to obtain specific indirect effect. I am afraid that multiplying two specific direct effects e.g. direct effect of A on M1 * direct effect of M1 on B is not sufficient to state existence of specific indirect effect of A on B via M1     (A → M1 → B) as we still need to test if this indirect effect is greater than zero in population.

Therefore, in my humble opinion we need to estimate a p-value for specific indirect effect or we need to conduct bootstrapping procedure to test if obtained specific indirect effect differs from zero (for details see Preacher & Hayes, 2008). My problem with AMOS comes with the fact that AMOS cannot give me any information about specific indirect effects. I can conduct bootstrap or obtain p-value for total indirect effect but not to separated ones.

I go through literature and I found an answer to this problem. I post it here as it might be useful for others with the same problem as mine.

For first, it seems that it is impossible to estimate specific indirect effect in multiple mediator models using AMOS:

 “Although accessible or user-friendly covariance structure modelling programmes, such as AMOS (Arbuckle, 2010) and EQS (Bentler, 2005), provide empirical significance tests of total indirect effects, they do not provide tests of specific indirect effects, which might be of primary concern to the applied educational researcher interested in examining specific intervening pathways in a postulated theoretical model (Mallinckrodt et al., 2006)” (Perera 2013, p. 40).

However, one can overcome these shortcomings of AMOS using so called phantom model approach proposed by Macho and Ledermann (2011):

“First, the phantom model approach may be regarded as the method of choice for those users employing the programs AMOS and EQS, which do not allow for the representation of specific effects by means of algebraic expressions (according to Jackson et al., 2009, these users make up about 50% of the the SEM users”  (Macho & Ledermann, 2011, p. 41).

The detailed procedure of using phantom model approach to estimate specific indirect effect is described in Macho and Ledermann (2011) and illustrative example can be found in Perera (2013).

References

Perera, H. N. (2013). A novel approach to estimating and testing specific mediation effects in educational research: explication and application of Macho and Ledermann’s (2011) phantom model approach. International Journal of Quantitative Research in Education, 1(1), 39. https://doi.org/10.1504/IJQRE.2013.055640

Preacher, K. J., & Hayes, A. F. (2008). Asymptotic and resampling strategies for assessing and comparing indirect effects in multiple mediator models. Behavior Research Methods, 40(3), 879–891. https://doi.org/10.3758/BRM.40.3.879

Macho, S., & Ledermann, T. (2011). Estimating, testing, and comparing specific effects in structural equation models: The phantom model approach. Psychological Methods, 16(1), 34–43. https://doi.org/10.1037/a0021763

http://quantpsy.org/pubs/preacher_hayes_2008b.pdf

http://www.academia.edu/download/32217569/Macho_Ledermann_2011_PM.pdf

Article A novel approach to estimating and testing specific mediatio...

Dear Konrad,

this link could help

http://amosdevelopment.com/features/user-defined/user-defined-general/index.html

Thank you Rebean!

User-defined estimates approach described under your link looks really promising. I hope it could solve my problem with specific indirect effects :)

I have found some additional information about user-defined estimates in AMOS here:

http://amosdevelopment.com/features/user-defined/user-defined-simple/Tutorial2/pdf/SimpleUserDefined2.pdf

I would like to ask if you possibly know any paper published in peer review journal /book applying user-defined estimates to test specific indirect effects in AMOS? Such a reference might be of particular importance as some authors postulate that this is impossible to test specific indirect effects in AMOS ( Perera 2013; Macho & Ledermann, 2011; see in my previous post)

Thank you one more time.

HI Konrad

You can find a way to test the effect of mediator variable …. in our paper:

A.S. H. Yousif a, Najm A. Najm a, Jasser A. Al-Ensour (2017). Total quality management (TQM), organizational characteristics and competitive advantage, Journal of Economic & Financial Studies, 05(04), 12-23 (Open access).

Thanks for sharing the information.

Hi Najm Najm,

Thank you for sharing you interesting paper, however I am afraid that your paper does not address directly my question.

I am not asking here about testing mediation but about testing specific indirect effects in multiple mediators model in Amos.

As far as I can ascertain all indirect relations in your models are always mediate only by one mediator variable, whereas my question concern situations in which you have more than one mediating variable and you want to estimate each indirect (or mediating) effect separately.

This problem might by solved by use of Amos user defined estimated as was describe in my previous post.

http://amosdevelopment.com/features/user-defined/user-defined-simple/Tutorial2/pdf/SimpleUserDefined2.pdf

Thank you on more time for sharing your research

As i understand you want calculate the specific indirect effect A on B while mediator1 inside model

i think there is a simple way for this. At first calculate model 1 without other variables. i added what i mean with model1

Then you would add other mediators and view other effect values

how we can calculate the p value of mediation effect in AMOS. anyone who knows about this.

Hello Zulqurain,

Possibly you might find helpful this video:

https://www.youtube.com/watch?v=j_yufPUjkwk

First there is a general introduction to a mediation concept and from about 6:50 the tutorial of how to test mediation in AMOS starts.

I guess you can constrain one of the indirect paths (1) to detect the indirect effect of (2) and vice versa to detect the effect of (1)

but I have a question, is it feasible in amos to provide the significance level of indirect effects without performing bootstrapping

Dear Manar Hamouda,

Thank you for your suggestion, as I wrote in my previous post, this problem might by solved by use of Amos user defined estimands. However placing a constrain of one of the indirect paths is interesting idea I will check it.

Answering your question, as far as I can ascertain you cannot obtain significance level for indirect effects in AMOS if you do not perform bootstrap because bootstrapping procedure is a means to get confidence intervals (CI) and p-values for test of indirect effects. Although in my opinion the most important aspect of indirect effect test is not p-value but test if CI does not include 0.

You might find useful this paper:

Preacher, K. J., & Hayes, A. F. (2008). Asymptotic and resampling strategies for assessing and comparing indirect effects in multiple mediator models. Behavior Research Methods, 40(3), 879–891. https://doi.org/10.3758/BRM.40.3.879

Thank you Konard, this was very useful. I hope that my suggestion can work with you well

Hi Konrad Kulikowski, HiKonrad, I have calculated specific indirect effects using user defined estimands in AMOS in a model with multiple mediators. Were you able to find any references reporting this finding?

Hi, Chitra Dey

If you would like to find a paper in the scope of your discipline using AMOS user-defined estimands you might try to search for “user-defined estimands” in Google Scholar, there are some papers that use this approach in different fields and possibly some of them, will be relevant to you.

In my view you might also refer to the Amos Development Corporation website where the use of AMOS user-defined estimands is described in details. http://amosdevelopment.com/features/user-defined/user-defined-simple/index.html

Also, James L. Arbuckle IBM® SPSS® Amos™ 25 User’s Guide contains a chapter about estimating indirect effect with user-defined estimands.

I know also at least one YouTube video showing this https://www.youtube.com/watch?v=9jGL45NuVAA possibly might serve as a reference.

If you would like to refer to a general idea of testing specific indirect effect with bootstrapping you might find useful a work of Hayes et al:

Hayes, A. F. (2009). Beyond Baron and Kenny: Statistical mediation analysis in the new millennium. Communication Monographs, 76, 408–420. doi:10.1080/03637750903310360

Preacher, K. J., & Hayes, A. F. (2008). Asymptotic and resampling strategies for assessing and comparing indirect effects in multiple mediator models. Behavior Research Methods, 40(3), 879–891. https://doi.org/10.3758/BRM.40.3.879

Hayes, A. F., & Scharkow, M. (2013). The relative trustworthiness of inferential tests of the indirect effect in statistical mediation analysis: Does method really matter?. Psychological science, 24(10), 1918-1927.

Thanks Konrad Kulikowski for the very detailed reply.

Hello all and especially #Chitra Dey and #Konrad Kulikowski, because my concern is somewhat similar to what you are recommending. Please help me with something -

I have 6 IVs, 4 Mediators and I DV in my model. All Medtrs. mediate the path between all 6 IVs and Dv. to calculate the specific indirect effects I created user-defined estimand by multiplying the reg weight of the path of IV1 to M1 and M1 to DV. But I am getting the syntax error "declaration expected"

Can you help me with this..I am not able to figure out what's the mistake that I am making in creating the estimand?

Hi Gunjan Sood,

It's hard to say from the distance, also I do not have longer access to AMOS as my current Uni. does not subscribe to this software - so I can be of a little help ;(

What I can advise you is inspecting this instruction videos:

http://amosdevelopment.com/features/user-defined/user-defined-general/index.html

and check if you are doing exactly the same things as they show.

Alternatively, you might try to use JASP software it is free and to my understanding, it is much easier to use with multiple mediators than AMOS, although it needs some initial learning.

https://jasp-stats.org/2018/07/03/how-to-perform-structural-equation-modeling-in-jasp/

good luck

Konrad

Thanks Konrad Kulikowski

Konrad Kulikowski did you find any peer review paper reference for user defined estimates to measure indirect effect of moderator?