Thanks for the video! What would be the difference between analyzing a control variable (for example, Gender Female /male) vs analyzing the model with groups (group 1: female, group 2 male) when to use one or the other process?
Control variables are trying to account for variance in your model in explaining the dependent variable. Two group analysis is examining if relationships are significantly different across the groups. Control variables are necessary if you think that a variable will change the outcome of your dependent variable....for instance Age and technology acceptance.
Great! I have a question regarding the measurement and structural models with control variables. When evaluating these models, should we include the control variables in the model and assess the model indices, or do we need to remove the control variables to assess the measurement and structural models?
I assess model fit before adding control variables. Many control variables will have relatively no influence but will cause unexplained variance in the model. In my opinion, assess the model fit with the initial structural model and then add you control variables after that. Is it wrong to assess model fit with the control variables included? No but it will usually produce more unexplained variance which could hurt model fit...especially if the control variables are not significant.
Sometimes a control variable (gender, age, education) is added as dummy, but I do not understand why it is done. Which one is correct? I am confused about it. By the way, could you send an example of control variable reported on the paper, especially with amos or smart? I do not know how to report. Thanks for your time.
If you have a control variable that is categorical, then you will need to dummy code that variable. If it is continuous, then no need for dummy coding. As for presenting results, I give a detailed breakdown in my book "Applied Structural Equation Modeling using Amos". Control variables are presented just like structural results. That is a little too complicated to present in the comments but I would encourage you to check out the discussion in the book. Hope this helps.
Mr. Collier, thanks for replying quickly. As you said, it is a little bit comlicated to comment. If I am stuck for presenting and commenting in the model, may I ask you?
You can add as many as you want. To be honest, I have never seen more than 10 control variables in a model and that seemed like a lot when I was reading through the article.
I enjoy your videos and your book "Applied Structural Equation Modeling Using AMOS" is very simple to follow, the language is very clear. Do you have videos or resources on SEM using R / Rstudio?
Hello sir, kindly answer to this question. You said that if the control variable does not make change in the beta value then we can remove it from the mode. But what if we keep it or show it in the model just to show that we controlled for it even if not significant?
It will be fine to leave it in the model. It is not having an effect but if you want to show a reader or reviewer that you controlled for the variable's effect, it will be fine to leave it in the model.
Hi, thank you very much! If we want to see if the relantionships between the obs variables -> mediator -> Y vary as a function of language, for example (let's say, a test in English and a test in Spanish, within subjects design). Should language be included as a covariate or as an obs variable?
@@joelcollier9387 Thank you for the reply! Even if I had a within-subject design? For example, the same group took the two tests (one in each language) . I just posted a question on Cross Validated with more details, the name is "How to design a within subjects structural equation model"
Hi, sir. I understand that we cannot use categorical variable as a control variable unless code it to binary variable. Just wonder can I include industry variable(1-19) as a control variable only show that effect of industry is controlled?
I'm not exactly sure what the industry variable is....if it is a scale of degree, then yes you can use it. If it is different categories of industries then you will have to dummy code it.
Dear Professor, how can we include ordinal or nominal variables as control variables (i.e. education with 8 categories ranged from 1 to 8) ; Do we need to re-code somehow our variables;
You can not treat a categorical variable (education ) as a continuous variable. So you would need to convert this to a categorical variable in the data. The easiest way is to dummy code the variable to high and low education levels (dichotomous) and control for the effects. If you are looking to control for more than two categories, then you are going to need to use a multicategorical approach. If that is the case, see my video titled "how to use multicategorical variables in SEM" Hope this helps.
No need for a control variable in a CFA. A control variable is to "Control" the influence of an independent to a dependent variable. With a CFA are you looking within and assessing validity. No need for a control variable.
No, I see no need in standardizing the data before running a SEM model. We do standardized the data when you create interaction terms with moderation but if it is a simply model, then there is no real need to standardize it first.
yes, they can be used. Categorical variables are a little more tricky. The easiest way is to set them up as a bimodal categorical variable. Otherwise, if it is multiicategorical it gets way more complicated. I have a video about how to use multicategorical variables if you want more information.
@@joelcollier9387 Is it right that there is a covariance between control variable error and independent error? Or no need to include errors? The control variable is affecting to independent and dependent variables. Thanks for your time again! :)
@@talzabidi1569 I always test it to all dependents because you are controlling for the effect in your total model and the influence of your dependent variables
Thanks a lot, Joel!
Thank you Prof. Joel!!
perfect as usual
Thanks for the video! What would be the difference between analyzing a control variable (for example, Gender Female /male) vs analyzing the model with groups (group 1: female, group 2 male) when to use one or the other process?
Control variables are trying to account for variance in your model in explaining the dependent variable. Two group analysis is examining if relationships are significantly different across the groups. Control variables are necessary if you think that a variable will change the outcome of your dependent variable....for instance Age and technology acceptance.
Great! I have a question regarding the measurement and structural models with control variables. When evaluating these models, should we include the control variables in the model and assess the model indices, or do we need to remove the control variables to assess the measurement and structural models?
I assess model fit before adding control variables. Many control variables will have relatively no influence but will cause unexplained variance in the model. In my opinion, assess the model fit with the initial structural model and then add you control variables after that. Is it wrong to assess model fit with the control variables included? No but it will usually produce more unexplained variance which could hurt model fit...especially if the control variables are not significant.
Sometimes a control variable (gender, age, education) is added as dummy, but I do not understand why it is done. Which one is correct? I am confused about it. By the way, could you send an example of control variable reported on the paper, especially with amos or smart? I do not know how to report. Thanks for your time.
If you have a control variable that is categorical, then you will need to dummy code that variable. If it is continuous, then no need for dummy coding. As for presenting results, I give a detailed breakdown in my book "Applied Structural Equation Modeling using Amos". Control variables are presented just like structural results. That is a little too complicated to present in the comments but I would encourage you to check out the discussion in the book. Hope this helps.
Mr. Collier, thanks for replying quickly. As you said, it is a little bit comlicated to comment. If I am stuck for presenting and commenting in the model, may I ask you?
How many control variables can you add to the model? What would be the max that you'd recommend?
You can add as many as you want. To be honest, I have never seen more than 10 control variables in a model and that seemed like a lot when I was reading through the article.
I enjoy your videos and your book "Applied Structural Equation Modeling Using AMOS" is very simple to follow, the language is very clear. Do you have videos or resources on SEM using R / Rstudio?
Thanks for supporting the channel and book. I do not have any videos on R. I am just sticking to AMOS right now.
Hello sir, kindly answer to this question. You said that if the control variable does not make change in the beta value then we can remove it from the mode. But what if we keep it or show it in the model just to show that we controlled for it even if not significant?
It will be fine to leave it in the model. It is not having an effect but if you want to show a reader or reviewer that you controlled for the variable's effect, it will be fine to leave it in the model.
Thank you very much sir for reverting. You reply would be help me a lot:)
Hi, thank you very much! If we want to see if the relantionships between the obs variables -> mediator -> Y vary as a function of language, for example (let's say, a test in English and a test in Spanish, within subjects design). Should language be included as a covariate or as an obs variable?
You would be better off testing language in a two group analysis so that you can see if the differences in language are significant.
@@joelcollier9387 Thank you for the reply! Even if I had a within-subject design? For example, the same group took the two tests (one in each language) . I just posted a question on Cross Validated with more details, the name is "How to design a within subjects structural equation model"
Hi does anyone know how to do exactly this but in R?
Hi, sir. I understand that we cannot use categorical variable as a control variable unless code it to binary variable. Just wonder can I include industry variable(1-19) as a control variable only show that effect of industry is controlled?
I'm not exactly sure what the industry variable is....if it is a scale of degree, then yes you can use it. If it is different categories of industries then you will have to dummy code it.
Dear Professor, how can we include ordinal or nominal variables as control variables (i.e. education with 8 categories ranged from 1 to 8) ; Do we need to re-code somehow our variables;
You can not treat a categorical variable (education ) as a continuous variable. So you would need to convert this to a categorical variable in the data. The easiest way is to dummy code the variable to high and low education levels (dichotomous) and control for the effects. If you are looking to control for more than two categories, then you are going to need to use a multicategorical approach. If that is the case, see my video titled "how to use multicategorical variables in SEM" Hope this helps.
Hello, Sir, Do we need to put control variables in CFA measurement model before structural model measurement ?
No, control variables are included to account for any bias in structural relationships. No need for a control variable in a CFA
Hi sir, do we need to re-consider the structural model fit again (e.g., CFI, TLI, etc) after adding the control variable ?
Yes, if you are going to include a control variable in the analysis, then model fit statistics need to include them.
can we use control variables for latent variable in measurement cfa model?
No need for a control variable in a CFA. A control variable is to "Control" the influence of an independent to a dependent variable. With a CFA are you looking within and assessing validity. No need for a control variable.
Hello, should I z-stardize my items before doing the SEM? Thank you!
No, I see no need in standardizing the data before running a SEM model. We do standardized the data when you create interaction terms with moderation but if it is a simply model, then there is no real need to standardize it first.
@@joelcollier9387 perfect, thank you so much!
Hello sir, Thanks for the useful video. Can the size of a firm and type of industry be control variables?
yes, they can be used. Categorical variables are a little more tricky. The easiest way is to set them up as a bimodal categorical variable. Otherwise, if it is multiicategorical it gets way more complicated. I have a video about how to use multicategorical variables if you want more information.
Thank you sir for the prompt reply. It would be great if you share that video for more clarity.
@@shaziazahid942 Here you go:
ruclips.net/video/VnmXfD6jjw8/видео.html
Hi, Prof. Can I assume control variable is affect to indepent variable(x) as well as mediator and dependent variable? Thank you for your time.
The control variables only have a relationship to the dependent variables. We are controlling for the influence in the outcome variables.
Thank you for your time. Do we allow to draw covariance arrow (bothside arrow) between independent variable and control variable…?
Please help!
@@yoonseoyang1977 That is correct! You will include a covariance between the control variable and independent variables
@@joelcollier9387 Is it right that there is a covariance between control variable error and independent error? Or no need to include errors? The control variable is affecting to independent and dependent variables. Thanks for your time again! :)
@@yoonseoyang1977 No need for error terms on the control variable and independent variables. Error terms are reserved for dependent variables.
Dr what about latent constructs , can we use it with latent construct full SEM?
Yes, Absolutely! The latent variable will have a relationship to the all the dependent variables in the model.
@@joelcollier9387 why do we need to make relationship to all dependents , instead to the ultimate dependent ?
@@talzabidi1569 I always test it to all dependents because you are controlling for the effect in your total model and the influence of your dependent variables