forum.smartpls.com

Thanks a lot Christian

It is helping me to clear my doubt.But still some doubts

when i captured EXP variable it was in continuous form (people entered 2.5 yrs, 3 years, 4 years and so on) and then i coded them into categories

If i m not wrong ,A variable can be treated as ordinal when its values represent categories with some intrinsic ranking.

Now here if my 1 indicates –very less experience (0-5yrs) , 2 indicates less experience(6-10 yrs) ,3 indicates medium experience (11-15 yrs) and 4 indicates high experience(more than 15).

Is it then looks like ordinal scale.

Now can I use it as a single latent variable EXP and then create interaction of this variable EXP and my other independent variable IV( measured on likert scale of 1-5) using "create moderating effect" in PLS

Please correct me if I wrong somewhere. Please clear my doubt

Another thing-
If I am using group comparison approach then my sample size for one group is - 100 cases and for other group sample size becomes 48. Can i model or use PLS for such a small sample size of 48. ( when I am having largest no of items as 6 in one construct/latent variable).

Thanks
Regards
Ruchi

I was just wondering why you reduce information by coding your metric/continous experience variable (years) into a ordinal scale of 4 categories.
Just using the years variable as a moderator will give you the most usefull insights: on how much the years of work influence the relationship between your two variables.

Thats true
I can do like this as well. But in case i dont capture it as continous and capture in ranges/categories, then how to go for this.

Please clarify me

Dear Sir

I have my EXP variable with values

Respondent 1 EXP (in years)
Respondent1 5.6
Respondent2 6
Respondent4 3
Respondent5 6
Respondent6 5.4
Respondent7 4.5
Respondent8 4.1
Respondent9 4
Respondent10 6.5
Respondent11 4
Respondent12 8.2
Respondent13 5.3

Now can I model it using smartPLS and "create moderating effect" using this data
For this first i should connect my LV ( with one indicator of EXP) to my dependent variable and then use "create moderating effect" on dependent variable with Moderator as-EXP and predictor as another IV( measured by 4 items)

Please correct me if i am wrong.

I am confused on this interaction effect modelling. It will be of great help if you can please clarify the things.

Thanks
Regards
Ruchi

Dear Prof. Ringle,

I just want to make sure about measuring moderating effect. I quote your statement as below :

" Generally sex is a moderator variable. In other words there is no other variable in the model having an impact on it. Furthermore we only consider the differences in the path coefficients if we use sex as a group variable in SmartPLS".

And I also quote others statementsA moderator was a variable that effects the direction and/or strength of the relation between an independent and a dependent variable (Baron and Kenny,1986).
The moderator effect would be said to be occurred if the relation was substantially reduced---from strong to a weak relation--- or increase --- from weak to strong relation (Baron and Kenny,1986).

“a moderating effect occurs when a third variable or construct changes the relationships between two related variables/ constructs. For example, we would say that a relationship is moderated by gender if we found that the relationship between two variable different significantly between males and females. For example, the relationships between two variables may be negative for males and positive for females or significant in one group and not the other” (Hair et al.,2010).

Considering the above statements I make conclusion that to see the moderating effect can be done by comparing the relationship between independent and dependent variables were not added by moderator and added by moderator or the relationship between independent and dependent variables of men group and women group if the moderator is gender.

Am I right with this conclusion?

Thanks
Best regards
Iin

Dear Prof. Ringle,

I just want to make sure about measuring moderating effect. I quote your statement as below :

" Generally sex is a moderator variable. In other words there is no other variable in the model having an impact on it. Furthermore we only consider the differences in the path coefficients if we use sex as a group variable in SmartPLS".

And I also quote others statements:

A moderator was a variable that effects the direction and/or strength of the relation between an independent and a dependent variable (Baron and Kenny,1986).
The moderator effect would be said to be occurred if the relation was substantially reduced---from strong to a weak relation--- or increase --- from weak to strong relation (Baron and Kenny,1986).

“a moderating effect occurs when a third variable or construct changes the relationships between two related variables/ constructs. For example, we would say that a relationship is moderated by gender if we found that the relationship between two variable different significantly between males and females. For example, the relationships between two variables may be negative for males and positive for females or significant in one group and not the other” (Hair et al.,2010).

Considering the above statements I make conclusion that to see the moderating effect can be done by comparing:

the relationship between independent and dependent variables were not added by moderator and added by moderator or

the relationship between independent and dependent variables of men group and women group if the moderator is gender.

I do not see the interaction effect/score of moderating variable to the dependent variable, but the result of the interaction effect on the relation of independent to dependent variable, I just see if there is the difference of the effect from independent to dependent variable of men and women group.

Am I right with this conclusion?

Thanks
Best regards
Iin[/quote]

Yes, and in other words...

Example using interaction term:

[1] Y = 5 + 2 * X + 1.5 * Gender – 2.5 * X * Gender

Where:
0 = male
1 = female

If the person is male:
[2] Y = 5 + 2 * X

If the person is female:
[3] Y = 5 + 2 * X + 1.5 * 1 – 2.5 * X * 1
[4] Y = 6.5 – 0.5 * X


If we use the multigroup analysis, we will have the equations [2] and [4] directly.

Best regards,

Bido

Dear Prof. Bido,

Thank you for your reply, Now I am quite sure about measuring moderating effect, but Still I want to make confirmation with you the result of my research, as below:

I find that:

The path coefficient from content (as independent variable) to behavioral intention (as dependent variable) without age (moderating variable) is 0.2172 (significance in p value < 0.05).

And the path coefficient from content (as independent variable) to behavioral intention (as dependent variable) with age (as moderating variable) is 0.2681 (not significance in p value < 0.05).

Since the first coefficient is significant and the second coefficient is not significant, I draw a conclusion that age moderating the effect of Content to Behavioral Intention.


Do you think that my conclusion is right?

Thank you

Best regards

iin

A comment from my side to your question also:

You can only say there is a moderating effect by age if you check the difference between this two paths.

Best regards,

Christian

Thank You for the answer.

I have another question.

I have read all the posts regarding moderating variable in this forum, I have conclusion that if the moderating variable is gender, it should be calculated by using multi group comparison. Is my conclusion right?

Is it right that when the moderating variable in nominal just like gender we cannot put it directly in the model?

Thank you

Iin

Thats absolutly right!
cn

Dear Christian,

Could you send me example of multigroup comparison by using gender as moderating variable?

Thank you

Regards
iin

Hey Iin,

if you send me your email I would can send you a example excel template for a t-test.

Best regards,

Christian

Dear sir

I would also like to see the excel for multigroup analysis by gender.

Can i send you my email

Thanks
ruchi

christian.nitzl wrote:Hey,

You should not use the variable/column with your dummy code in your model. After you splitting your file in the different groups there are only e. g. 1 in the column. Therefore you have got an error message. (A second reason could be, that your model is not recursive.)

Here a very good article, this time in English, who also perform a group comparison with four groups in PLS:

Eberl, Markus (2010): An Application of PLS in Multi-Group Analysis, in: Esposito Vinzi, V./Chin, W./Henseler. J./Wang, H. (Eds.): Handbook of Partial Least Squares: Concepts, Methods and Applications, Berlin, pp. 487-514.

Best regards,

Christian

Hi Christian and everybody...

I have got this article. Could you or somebody explain to me about how to calculate the "diff", "pooled s.e" and "p" as written at page 507 Vinzi et al book.

Cheers,