Hello Jyotsna,

The more general approach of path analysis (a subset of SEM analysis in which all variables are manifest/observed, with no reciprocal influences) could easily accommodate just about any model you could contrive with your set of variables. For that matter, ordinary least squares regression analysis could address any question regarding whether a given path connecting two variables was non-zero.

There are numerous SEM programs (including ones that are free, such as the lavaan library in R) from which to choose.

As you haven't specified your research question(s) of interest, it's hard to know whether a specific, pre-defined model in the PROCESS macro would evaluate it.

Good luck with your work.

Moderate mediation models as I have seen present mostly one dependent variable, but with path analysis you must include variables created with interactive effect of the moderating variable to report the moderate mediation index (I recommend the free JAMOVI program), another way are the PLS-SEM models that also fit the model you refer to.

https://www.jamovi.org/release-notes.html

Hello David Morse ,

thank you for sharing your views on my query. I am working on understanding the effectiveness of social media advertising on working women' buying behaviour, where I have:

1. WW online buying behaviour as Dependent variable

2. personalised SMAds and Purchase intentions as Independent variables.

3. Ad innovativeness, Ad usefulness, and transaction medium are the Mediating variables.

4. Risk being a moderating variable (Hayes &Preachers model 1 for moderation)

I have read and followed articles on mediation and moderation, where researchers have applied regression analysis using Hayes and Preacher' approach to run mediation and moderation effect. (Hayes, A. F. (2012). PROCESS: A versatile computational tool for observed variable mediation, moderation,and conditional process modeling [White paper]. Retrieved from http://www.afhayes.com/public/process2012.pdf).

i am in a bit of confusion regarding how to go further, suggestions have come to follow AMOS but I am yet not able to find the appropriate pieces of literature which could guide me further with my concerned area.

Hello again Jyotsna,

The basic model, given the information in your second post, would imply that:

1. The direct links below would be weak (relative to others) or zero:

a. SM -> BB (SM ads to online buying behavior)

b. PI -> BB (Purchase intentions to BB)

c. Risk -> BB

2. The direct links below would be strong(er) and non-zero:

a. SM -> AI (SM ads to Ad innovation)

b. SM -> AU (SM ads to Ad usefulness)

c. SM -> TM (SM ads to transaction medium)

d. PI -> AI

e. PI -> AU

f. PI -> TM

g. AI -> BB

h. AU -> BB

i. TM -> BB

3. The moderation implies you'd want to look at these links

a. SM * Risk -> AI

b. SM * Risk -> AU

c. SM * Risk -> TM

d. PI * Risk -> AI

e. PI * Risk -> AU

f. PI * Risk -> TM

g. SM * Risk -> BB

h. PI * RIsk -> BB

If the last two (3g, 3h) are strong relative to 3a-3f, then the moderation isn't mediated. If weak relatively, then the moderation is mediated. As well, the moderation links should be non-zero, and ideally strong or stronger than the direct links involving SM and PI to support the claim that Risk is a moderator.

All of these can easily be set up in AMOS, LISREL, EQS, CALIS, lavaan, or a host of other SEM software packages.

Good luck with your work.

The conceptual model is not clear. I wish it was shared for further discussion. Almost have a similar challenge where my two independent variables link to one dependent variable. So am wondering as to test moderation effect on for each one to one relationship in SmartPls or to add the two moderating terms and run the whole model at once.