Dear Crowson, first of all, I would like to say that your channel has been a great blessing for young researchers. I greatly benefit from your videos and especially your ppt presentations. I'm really grateful to you. Is it possible for you to upload a video about mediation/moderation variable analysis using covariate or control variables?
Thank you for your encouragement with my work. I'll see what I can do on that (I still have a number of projects in the cue ahead). However, in the meantime, just remember that you can add covariates into the Covariates box when you perform mediation or moderation analysis. Mathematically, the covariates are included as control variables in all regression equations involving prediction of an endogenous variable (which will be any mediators and the consequent) in the model. Thank you again!
concerning your process output and the odd columns on min 18: Just change the font in spss output to one where every letter has the same width. Then all columns are perfectly in line. :) Had the same issue and could solve it that way.
Dear Sir, as a young researcher i found your video very helpful!Thank you very much!i would like to ask about more reference about Process macro but in a practical way. and something else, how i would interpret the mediation if my low and uper numbers are statistically improtant but negative?
With Process model 4, you can specify your X variable as a multicategorical variable. You can also specify your Y variable as a binary variable [If you do this, the second regression model becomes a binary logistic regression]. However, the mediators must be continuous.
Thank you for the video! It was really informative and helped me a lot. I have 2 questions regarding a model with 2 mediators though: 1) If I want to compare which path (a or b) has a stronger impact on DV, then I just have to test if the indirect effects are stat different and compare them, right? 2) In my case I have two IVs. Can I use Hayes Process for that? Thanks a lot!
Hey! Thanks a lot. What if my I.V is binary with two levels (coded as 1 and 0). What is the interpretation if a full mediation occurs? All other variables are continuous (covariate, mediator and D.V).
Thank you for your explanation, I learnt a lot. But how to account for more independent variables in the model? From what I can see in the video, we can only have one IV under X variable. Thank you!
Process is set up for a single X and a single Y. However, you could 'trick' Process into testing mediation using multiple X's by running a series of models with one X as IV and the remaining X's as covariates. The indirect effect in each model would be associated with the X treated as the IV in each model. Alternatively, you might run path analysis using an SEM program. Hope that helps. Cheers!
Hello Mike Please i need clarification on whether i need to do a linear regression analysis first on SPSS for variable X-Y and X-M before using HAYES process model to do the mediation analysis or should i just do a midiation analysis using HAYES process model only and get the results for X-Y and X-M from the process model results summary ?
The paths in the mediation model are estimated using separate ols regressions. There's no need to estimate them using the regression menu in Spss since they are already provided in the process output. Cheers !
If I want to report the standardized effects in my thesis, can I then use the significance level of the non-standardized effect for the standardized effects? I ask because no p-values are given for the standardized effects
The p-values for the standardized effects in the individual regression models would be the same as those for the unstandardized effects. I'm not a huge fan of only reporting standardized effects and p-values [I tend to report both unstandardized and standardized coefficients, but that's my preference]. The p-values for the unstandardized and standardized effects are the same for the regressions. Keep in mind that when you select Standardized effects when using Process Model 4, it will also print out standardized indirect effects (at the end of the output) along with bootstrapped confidence intervals for those effects. [You will see this appear below the output for unstandardized indirect effects and bootstrap confidence intervals] You won't get p-values for any of these, but the confidence interval can be used for testing significance of any indirect effects you compute. If 0 falls outside the lower and upper bound for an interval, then the indirect effect is deemed as statistically significant. You can report on the bootstrap confidence intervals along with for your indirect effect and this will convey information about your estimated indirect effect and significance.
Dear Crowson, first of all, I would like to say that your channel has been a great blessing for young researchers. I greatly benefit from your videos and especially your ppt presentations. I'm really grateful to you.
Is it possible for you to upload a video about mediation/moderation variable analysis using covariate or control variables?
Thank you for your encouragement with my work. I'll see what I can do on that (I still have a number of projects in the cue ahead). However, in the meantime, just remember that you can add covariates into the Covariates box when you perform mediation or moderation analysis. Mathematically, the covariates are included as control variables in all regression equations involving prediction of an endogenous variable (which will be any mediators and the consequent) in the model. Thank you again!
You are a legend. Thank you for this video
You are very welcome! Thanks for dropping by :)
THANK YOU SO MUCH!!!
concerning your process output and the odd columns on min 18: Just change the font in spss output to one where every letter has the same width. Then all columns are perfectly in line. :) Had the same issue and could solve it that way.
I appreciate it. Thanks for the tip :)
"Courier new" worked for me :)
Thank you!!!
Thanks a lot!
Thanks a lot ❤
Dear Sir, as a young researcher i found your video very helpful!Thank you very much!i would like to ask about more reference about Process macro but in a practical way. and something else, how i would interpret the mediation if my low and uper numbers are statistically improtant but negative?
Thank you very much for the explanation. One question, can I use the analysis with all categorical data?
With Process model 4, you can specify your X variable as a multicategorical variable. You can also specify your Y variable as a binary variable [If you do this, the second regression model becomes a binary logistic regression]. However, the mediators must be continuous.
Thank you for the video! It was really informative and helped me a lot. I have 2 questions regarding a model with 2 mediators though:
1) If I want to compare which path (a or b) has a stronger impact on DV, then I just have to test if the indirect effects are stat different and compare them, right?
2) In my case I have two IVs. Can I use Hayes Process for that? Thanks a lot!
Hey! Thanks a lot. What if my I.V is binary with two levels (coded as 1 and 0). What is the interpretation if a full mediation occurs? All other variables are continuous (covariate, mediator and D.V).
Thank you for your explanation, I learnt a lot. But how to account for more independent variables in the model? From what I can see in the video, we can only have one IV under X variable. Thank you!
Process is set up for a single X and a single Y. However, you could 'trick' Process into testing mediation using multiple X's by running a series of models with one X as IV and the remaining X's as covariates. The indirect effect in each model would be associated with the X treated as the IV in each model. Alternatively, you might run path analysis using an SEM program. Hope that helps. Cheers!
Hello Mike
Please i need clarification on whether i need to do a linear regression analysis first on SPSS for variable X-Y and X-M before using HAYES process model to do the mediation analysis
or should i just do a midiation analysis using HAYES process model only and get the results for X-Y and X-M from the process model results summary ?
The paths in the mediation model are estimated using separate ols regressions. There's no need to estimate them using the regression menu in Spss since they are already provided in the process output. Cheers !
If I want to report the standardized effects in my thesis, can I then use the significance level of the non-standardized effect for the standardized effects? I ask because no p-values are given for the standardized effects
The p-values for the standardized effects in the individual regression models would be the same as those for the unstandardized effects. I'm not a huge fan of only reporting standardized effects and p-values [I tend to report both unstandardized and standardized coefficients, but that's my preference]. The p-values for the unstandardized and standardized effects are the same for the regressions.
Keep in mind that when you select Standardized effects when using Process Model 4, it will also print out standardized indirect effects (at the end of the output) along with bootstrapped confidence intervals for those effects. [You will see this appear below the output for unstandardized indirect effects and bootstrap confidence intervals] You won't get p-values for any of these, but the confidence interval can be used for testing significance of any indirect effects you compute. If 0 falls outside the lower and upper bound for an interval, then the indirect effect is deemed as statistically significant. You can report on the bootstrap confidence intervals along with for your indirect effect and this will convey information about your estimated indirect effect and significance.