For example sex, age or level of studies.
Thanks!!
How can I include control variables in smartpls?
I remembered Dr. Bido once mentioned two methods. I do it in this way:
If your variable have n levels, such as high, mediem, low. You need to use n-1 dummy code variables to represent. For example, use one dummy code variable to represent gender, Male-1 for male and 0 for female.
For catergorized variable such as age, for example 16-20, 21-25……I think we need to use the median to represent each scale. For example, if the respondant chose 21-25, it will be replaced by 23. Some other variables like company size also need to be treated like this.
Other method maybe is group anlysis? You can search for it.
If your variable have n levels, such as high, mediem, low. You need to use n-1 dummy code variables to represent. For example, use one dummy code variable to represent gender, Male-1 for male and 0 for female.
For catergorized variable such as age, for example 16-20, 21-25……I think we need to use the median to represent each scale. For example, if the respondant chose 21-25, it will be replaced by 23. Some other variables like company size also need to be treated like this.
Other method maybe is group anlysis? You can search for it.
Do the formative by using PLS!!
- Diogenes
- PLS Super-Expert
- Posts: 899
- Joined: Sat Oct 15, 2005 5:13 pm
- Real name and title:
- Location: São Paulo - BRAZIL
- Contact:
Hi,
If you just want to remove the effect of the control variables:
- create (n-1) dummy for the n categories of the nominal control variable.
- include in the model a formative LV with these dummy
- run a model just with the control variable, run the complete model, compare the R2 chance to assess the contribution of the predictors.
A didactical example is available at:
FALK, R. F.; MILLER, N. B. A Primer for Soft Modeling. Ohio: The University of Akron Press, 1992.
If you want to assess the results for each level of the nominal control variable, the best way is the multi-group analysis.
Best regards,
Bido
If you just want to remove the effect of the control variables:
- create (n-1) dummy for the n categories of the nominal control variable.
- include in the model a formative LV with these dummy
- run a model just with the control variable, run the complete model, compare the R2 chance to assess the contribution of the predictors.
A didactical example is available at:
FALK, R. F.; MILLER, N. B. A Primer for Soft Modeling. Ohio: The University of Akron Press, 1992.
If you want to assess the results for each level of the nominal control variable, the best way is the multi-group analysis.
Best regards,
Bido
Dear Sir
In my model my main dependent variable is behavioral intention i..e BI. and there are 4 variables(Var1, var2 , var3 and Var 4) which effects this BI. The Var1, var2 variables are also endogenous variables i..e they are getting effected by 2 variables each
Var1, var2 , var3 and Var 4 are directly linked to BI (an arrow pointing towards BI)
varX, varY both are directly linked to Var 1 and var 2.
Now for controlling for AGE and GENDER, I should connect the link to BI only or to all the endogenous variables in the model (i..e BI, var1 and Var2)
Thanks
Ruchi
In my model my main dependent variable is behavioral intention i..e BI. and there are 4 variables(Var1, var2 , var3 and Var 4) which effects this BI. The Var1, var2 variables are also endogenous variables i..e they are getting effected by 2 variables each
Var1, var2 , var3 and Var 4 are directly linked to BI (an arrow pointing towards BI)
varX, varY both are directly linked to Var 1 and var 2.
Now for controlling for AGE and GENDER, I should connect the link to BI only or to all the endogenous variables in the model (i..e BI, var1 and Var2)
Thanks
Ruchi
- Diogenes
- PLS Super-Expert
- Posts: 899
- Joined: Sat Oct 15, 2005 5:13 pm
- Real name and title:
- Location: São Paulo - BRAZIL
- Contact:
Hi,
A “control variable” is a variable that affects the dependent variable, but it is not used as independent variable.
If we do not include the control variables, the relation between ID and DV should be underestimated or overestimated.
For these reasons I think that you should connect the control variable just to the dependent variables.
Best regards,
Bido
A “control variable” is a variable that affects the dependent variable, but it is not used as independent variable.
If we do not include the control variables, the relation between ID and DV should be underestimated or overestimated.
For these reasons I think that you should connect the control variable just to the dependent variables.
Best regards,
Bido
Dear Sir
Thank you for the answer.
When i ran my model with the control variable (age), most of the paths are significant (all same as in the model without control variable ) except one path. Also the path of age to the dependent variable (DV) is significant. It means that there is direct effect of age on DV. Am i right ?
It means that all the paths which are significant are regardless of age whatever the age may be these significant paths will be significant.
Let me know if I am wrong
Now i to do multigroup analysis i divided age into 3 groups ( with sample size of each group as 380, 59 and 87) and ran the same model ( i.e without control variable- age) for these 3 groups.
Now i found that majority of the paths are significant in first group , few in second group and only one in third group )
I have a doubt that if i m controlling for age and the paths are significant in the overall model with control variable then why the paths are becoming insignificant in the third age group subsample.?
Thanks
Ruchi
Thank you for the answer.
When i ran my model with the control variable (age), most of the paths are significant (all same as in the model without control variable ) except one path. Also the path of age to the dependent variable (DV) is significant. It means that there is direct effect of age on DV. Am i right ?
It means that all the paths which are significant are regardless of age whatever the age may be these significant paths will be significant.
Let me know if I am wrong
Now i to do multigroup analysis i divided age into 3 groups ( with sample size of each group as 380, 59 and 87) and ran the same model ( i.e without control variable- age) for these 3 groups.
Now i found that majority of the paths are significant in first group , few in second group and only one in third group )
I have a doubt that if i m controlling for age and the paths are significant in the overall model with control variable then why the paths are becoming insignificant in the third age group subsample.?
Thanks
Ruchi