Thank you for your video. How will be a moderated regression analysis with 4 independent variables moderated by one moderator variable to predict a single dependent variable?
Thanks for your video, Derek. In time 4:52, it seems a mistake: you inputted Gender at Process M variable box instead of Proposed moderator W. Is it correct?
For the old version of Process is correct in the new one it depends .. if you follow Preacher and Hayes process publication they actually tell you where to put the variable depending on what you are trying to find out
Thank you for the video! How shall I analyse if gender has 3 categories (+diversity) and I have only one person who is diverse? Shall I a) exclude the person from the analysis as the category did not reach above 30? or b) create dummy variables? If b) how?
Hi! Thanks for ur video it is really helpful. I have a question here. I am using age as my moderating variables and I will put them into two groups which are below 35 years old (1st group) and above 35 years old (2nd group). 1.Could you advice which type of measure for this? Is it ordinal or scale?
Hi Dr Derek, I would like to ask one question. If I group below 35 years old become young and code as 0, and above 35 years old become older group code as 1, should I mean-centred this variable before conducting the regression analysis using macro process?
Hi! Thank you for the video. I have a question. I have a moderated moderation model (model 3) with a continuous DV, a nominal/dichotomous IV and two nominal/dichotomous moderators. Under Options, should I select mean centering or not? Thank you!
Hi Dr. Ong, your video is very helpful, just wondering do you have the video on how to reporting the categorical moderation analysis (a standard format ) THANKS
Tongtongg Hou greetings I am not sure what you mean by a standard format....although the format I have mentioned here is just a suggestion...there are any other formats available it all depends on the requirements of your research or the journal your submitting to
Morning Dr. Ong, I totally agree with your formate provided above! I would like to know what is the best way to report the results in the dissertation (say in the multiple regression tests, you need to state the beta value, R square change, and p-value blablabla) , THANKS
Tongtongg Hou p-value is not really necessary if you are indicating the asterix on the beta and add it as a subscript at the bottom of the table...Do check out the video on multiple regression and how best to report the results as I have a suggested format ...cheers :)
THANK YOU DEREK!! And thank you for the speedy reply. Got PROCESS :-) Please could you also advice, would the outcome be violated if I run regression with non-parametric data? If so, what other ways could I do apart from removing outliers and adjusting outliers? (I have already tried these two options) Thank you!!
Evauk88 its really hard to say.....you might want to remove multicollineared variables as well....but usually if it is social science data and you have enough data you should not violate the normality assumption
Dear Derek, Thank you for the great instruction. I have for the moderators age and income as scale variable. Do i need to make dummy variables? Or can i just do what you did, one column with all the three different categories of age (Millenials, Generation Y and babyboome) and label them as 1,2 and 3 respectively.
Hello Professor, thanks for your VDO. I have 3 independent variables, 2 moderators (arousal [measured by 5 point Likert scale] and income variable (below 15K THB, 15K-20K, 20K-30K, 30K-40K,40K-50K, 50K above)) and 1 dependent variable. 1. Do I have to code income into 5 dummy variables before running regression? 2. When testing hypothesis, do i have to input every independent variable in IV box or I have to put it based on my hypotheses? For example, place identity has a positive effect on purchase intention is my first hypotheses. To test this, can I run one big data for regression or I should put only place identity in independent and purchase intention in dependent? sorry for very long questions.
thankyou Derek for the quick reply. so in order to get the moderator variable for age should the 1, 2 and 3 be multiplied with the main independent variable?
Hi Dr Ong! Do we need to click anything in particular to specify that it is a dichotomous moderator? Process does not seem to detect that it is dichotomous
if I could to ask ... I am dealing with two independent variables, one dependent variable, and one moderator which consists of three subgroups. the question is can I separate the sample for these three subgroups and do the linear multiple regression as a moderating analysis? or what is the best way to deal with this problem? thank you for reading this comment or thinking about it I appreciate your efforts in helping others Anoud
I have yet to explore multicategorical moderation at this point. You may do the individual regressions using the subsamples provided that the subsamples have enough sample sizes for the analysis ..cheers
I am so grateful for your response. the total sample I have is 861. The three subsamples (179/ 309/ 373). Is it ok to to do the individual regressions using the subsamples?
Hi Dr. Ong, is it possible to run moderation analysis with PROCESS if my independent and moderator variables both are dichotomous? and if it possible, I would appreciate if you can share us how to interpret the results. Thank you.
Hi :) I have exactly the same problem. My independet and my moderator variable ar both dichotomous. My dependet variable is metric. Is this working like in the video? Thx in advance!!
Thank you for your video! It helps me a lot in doing my theses. I have a few questions to ask: 1. Is the p-value of the interaction below 0.05 to be considered significant? My 'coeff' in the interaction column shows '-.9701' and the p-value of the interaction is '.0048'. Is this considered significant? 2. I'd watched your video about the continuous moderator analysis as well. My second moderator is the age of the participants. I remain it as a continuous variable and run the analysis. The 'coeff' in the interaction column shows '.0928' and the p-value for the interaction is '.000'. Is this considered significant? If it is significant, how should I generate the graph? I've used the same method you taught in this video, but the spss doesn't allow me to set the marker by using a continuous data. Should I group my data and make it become a categorical variable? 3. Are there any assumptions that need to be tested before carrying out this two analysis? Thank you so much. Hope to hear from you soon.
1. yes it is significant - 2. Yes it is significant.... try to set the markers in the data to draw your graph to see the changes it would be easier that way Cheers and all the best
Dear Derek, Further to Sophie Ho's question on age as scale variable, we have categorised age as millennial, Gen Y and Baby boomers and our respondents have selected their respective category. In order to apply this variable for moderator effect (with 3 categories) what steps should I follow? would appreciate if you could guide us through. Best regards Jyothika Shetty
Unfortunately I have not been able to venture into more than 2 categories.. so instead I create an interaction variable for moderator and then run a linear regression to see if the interaction is significant ( if so then there is moderation effect)
HI Dr.Ong. Thank you for the video, it's quite helpful. Could I get some direction from yourself on the following scenario: I'm planning to conduct a study that'll have 2 categorical independent* variables with 2 levels each, 1 moderator variable , and 1 continuous dependent variable. The results of my moderator (which will most likely be measured on a scale of 1 to 7) are to be scored as low (1-2.9), medium (3-5), and high (5.1-7). Is it possible to use the PROCESS macro in this case wherein my moderator will have 3 'levels' i.e. low, medium, and high? How should I assign them values if there are 3 'levels' with PROCESS? Thank you in advance. Note: I am still undecided on whether I will propose my variables as being correlated or as predictor/criterion variables.
in SPSS 25 goto Extensions, scroll down to Utilities, then Install Custom Dialog (Compatibility model). Locate 'process.spd' in the Custom dialog builder folder. Select it, it will ask if you want to Install, and, of course you do. Let me know.
Hii Derek, thank you for the video, it's extremely helpful! But- I'm currently using SPSS version 23, in the Analyze > Regression, there's NO option for PROCESS. What can I do??? Is there an alternative way to measure the independent, dependent and gender as you've presented in this video??? Thank you!
THANK YOU DEREK!! And thank you for the speedy reply. Got PROCESS :-) Please could you also advice, would the outcome be violated if I run regression with non-parametric data? If so, what other ways could I do apart from removing outliers and adjusting outliers? (I have already tried these two options) Thank you!!
Hello, Great video- thank you for putting this together. One quick question: Does PROCESS automatically centralize your continuous IV, or do you still need to do that before conducting the moderated regression?
@@tyriqueking8873 treating it as categorical variable would mean to treat it as multicategorical .. for that I am not entirely sure how to analyse it in Process
Hello Derek! Thanks for this great presentation! I would like to ask you if it is possible to share the paper used in this example. Which scale dou you use in order to measure "Perceived Value" and "Product Design"? Once more, thank you!
hello Derek and thank you for this demonstration, but does this methode work if the moderator have more than 2 categories? would you please answer me because i need it ASAP
Yes it does although the calculation would be a little bit more complicated.. I havent had the time to fully engage on its complexities so i would like to leave you with this paper which is very useful and does a simulation of multicategorical moderation Andrew F. Hayes & Amanda K. Montoya Pages 1-30 | Published online: 18 Jan 2017 A Tutorial on Testing, Visualizing, and Probing an Interaction Involving a Multicategorical Variable in Linear Regression Analysis Communications Methods and Measures Journal, 11(1), pp.1-30 Taylor and Francis doi.org/10.1080/19312458.2016.1271116 www.tandfonline.com/doi/full/10.1080/19312458.2016.1271116?src=recsys All the best
Derek Ong thank you very much for your fast reply, but I have problems at it's application it says error even when I used it with gender (i used 0, 1 codation)
oh i'm so sorry Derek, i'm using the latest version of PROCESS v3.0, and at the place of the moderator variable here in your version there is at mine mediator var..that's why it's giving me an Error.. but still, should I put my moderator at the moderator window W or Z below because I've got really confused!!!
hi sir. i first want to be thankful to you for this video as it helped me in doing my thesis. i have a question that i have used gender as moderator and beta values are negative for gender and for interaction , both are significant . i want to ask how to interpret these negative beta values for gender and interaction.
Nida Raja greetings it just basically means that it depends what you coded 0 and 1 as.....meaning that the relationship between your gender and dependant variable is stronger for the lower number than with the higher number...
At this stage, Hayes has not showed how to do this online with multicategorical moderator. However he has mentioned that the new version and tutorials will be able to do this.. unless you would like to explore the possibility of using dummy variables for comparisons you want to review this paper www.tandfonline.com/doi/full/10.1080/19312458.2016.1271116. Hope this helps cheers
Okay! Can the conditional interaction be significant for all of the categories but the p values above are non significant?! Does it look logic to you?! Not 1, not 2 all if them!!
if you don;t mind... you may send me your data ( only send me your independent, dependant and moderator variable and let me try it for you to see what is the problem
Derek Ong good morning Derek..thank you very much, it worked finally, it seems that after doing the reg for other multicats var the model had changed and I wasn't paying attention to it, once I choosed model 1 it worked! It was a crazy day yesterday
Thanks a lot for you step by step walkthrough. Really helped me reaching a good analysis level during my master thesis.
Thanks and cheers....All the best
Thank you very much...very good explain ....👍👍👍
Hello, thanks for your great presentation, but can you instruct us how to report the results in APA style?
Thank you for your video. How will be a moderated regression analysis with 4 independent variables moderated by one moderator variable to predict a single dependent variable?
Hi thank you for your question .. you can do each relationship and test the moderator one by one
Thanks for your video, Derek. In time 4:52, it seems a mistake: you inputted Gender at Process M variable box instead of Proposed moderator W. Is it correct?
For the old version of Process is correct in the new one it depends .. if you follow Preacher and Hayes process publication they actually tell you where to put the variable depending on what you are trying to find out
@@derekong1609 Thanks!
Hello! How would the moderation work with a categorical independent variable (X) and continuous moderator and dependent variable?
Choose one of the other in seperate analysis
@@derekong1609 thanks for getting back to me! one of the other what?
@@karlaveras1793 meaning do each of the moderation one by one separately....my apologies :)
Hi Derek, very helpful video, do you have more information about how to write up and interpret these results? thank you
Hi Yes ...it is all in the video as well ...
Thank you for the video! How shall I analyse if gender has 3 categories (+diversity) and I have only one person who is diverse? Shall I a) exclude the person from the analysis as the category did not reach above 30? or b) create dummy variables? If b) how?
Greetings you can either exclude it or if want to combine this persyn into one of the other categories and label it accordingly :)
@@derekong1609 perfect, thank you :)
Hi! Thanks for ur video it is really helpful. I have a question here. I am using age as my moderating variables and I will put them into two groups which are below 35 years old (1st group) and above 35 years old (2nd group).
1.Could you advice which type of measure for this? Is it ordinal or scale?
Hi you can use nominal for your group (young vs old)
Thank you for your response! I will just put them as nominal for the measure 👍
Hi Dr Derek, I would like to ask one question. If I group below 35 years old become young and code as 0, and above 35 years old become older group code as 1, should I mean-centred this variable before conducting the regression analysis using macro process?
@@foowuijia8592 you do not need to mean center the moderator if it is categorical
@@derekong1609 Thank you!
Hii Derek, very informative video but may I know how to get PROCESS Andrew F. Hayers in the SPSS because I don't have that part in my SPSS
You will need to download it online form Precher and Hayes website
@@derekong1609 I looked into it but it's not free to download anymore 🥹 do u have other ways?
@@NurHidayah-op6vf you may want to explore the Baron and Kenny method
@@derekong1609 Okay, I'll try this first. Thank you so much Derek! You're very helpful, and so are your videos 👍
@@NurHidayah-op6vf I have updated my video details on the link you can use to download PROCESS Macro and it still works.
Can you clarify that if you have a mixture of categorical and continuous covariates that they are all placed in the covariate box. Thanks
Greetings Catherine... usually the covariate would be a continuous variable...
Greetings covariates are used for control variables
Hi! Thank you for the video. I have a question. I have a moderated moderation model (model 3) with a continuous DV, a nominal/dichotomous IV and two nominal/dichotomous moderators. Under Options, should I select mean centering or not? Thank you!
If I am using gender as females -1 and males - 2, do I still need to dummy code it into 0 and 1?
It would be better to do so all you need to do is just recode them in SPSS i show how this is done in one of my videos
Hi Dr. Ong, your video is very helpful, just wondering do you have the video on how to reporting the categorical moderation analysis (a standard format ) THANKS
Tongtongg Hou greetings I am not sure what you mean by a standard format....although the format I have mentioned here is just a suggestion...there are any other formats available it all depends on the requirements of your research or the journal your submitting to
Morning Dr. Ong, I totally agree with your formate provided above! I would like to know what is the best way to report the results in the dissertation (say in the multiple regression tests, you need to state the beta value, R square change, and p-value blablabla) , THANKS
Tongtongg Hou p-value is not really necessary if you are indicating the asterix on the beta and add it as a subscript at the bottom of the table...Do check out the video on multiple regression and how best to report the results as I have a suggested format ...cheers :)
THANK YOU DEREK!! And thank you for the speedy reply. Got PROCESS :-)
Please could you also advice,
would the outcome be violated if I run regression with non-parametric data?
If so, what other ways could I do apart from removing outliers and adjusting outliers?
(I have already tried these two options)
Thank you!!
Evauk88 its really hard to say.....you might want to remove multicollineared variables as well....but usually if it is social science data and you have enough data you should not violate the normality assumption
Dear Derek, Thank you for the great instruction. I have for the moderators age and income as scale variable. Do i need to make dummy variables? Or can i just do what you did, one column with all the three different categories of age (Millenials, Generation Y and babyboome) and label them as 1,2 and 3 respectively.
Hi Sophie... you don't have to do that if your scale are sufficient... you can just use the Continuous Moderator for this purpose
Hello Professor, thanks for your VDO. I have 3 independent variables, 2 moderators (arousal [measured by 5 point Likert scale] and income variable (below 15K THB, 15K-20K, 20K-30K, 30K-40K,40K-50K, 50K above)) and 1 dependent variable.
1. Do I have to code income into 5 dummy variables before running regression?
2. When testing hypothesis, do i have to input every independent variable in IV box or I have to put it based on my hypotheses? For example, place identity has a positive effect on purchase intention is my first hypotheses. To test this, can I run one big data for regression or I should put only place identity in independent and purchase intention in dependent?
sorry for very long questions.
Hi surangnut... I hope I addressed this question
thankyou Derek for the quick reply. so in order to get the moderator variable for age should the 1, 2 and 3 be multiplied with the main independent variable?
Yes... use the Baron and Kenny Method... :)
Hi Dr Ong! Do we need to click anything in particular to specify that it is a dichotomous moderator? Process does not seem to detect that it is dichotomous
Just make sure that the coding for the categorical variable is coded 1 and 0
Dr, if the moderator is one of the variables and not categorical, what type of graph should choose? instead of group scattered.
break it up into 3 separate groups if you want to graph it low medium and high
It would be great if you show the step in you tube Dr.
Please watch video 6 transformation of quantitative variables
if I could to ask ... I am dealing with two independent variables, one dependent variable, and one moderator which consists of three subgroups.
the question is can I separate the sample for these three subgroups and do the linear multiple regression as a moderating analysis?
or what is the best way to deal with this problem?
thank you for reading this comment or thinking about it
I appreciate your efforts in helping others
Anoud
I have yet to explore multicategorical moderation at this point. You may do the individual regressions using the subsamples provided that the subsamples have enough sample sizes for the analysis ..cheers
I am so grateful for your response. the total sample I have is 861. The three subsamples (179/ 309/ 373). Is it ok to to do the individual regressions using the subsamples?
I think it should be enough
Thank you so much 🌹
Hi Dr. Ong, is it possible to run moderation analysis with PROCESS if my independent and moderator variables both are dichotomous? and if it possible, I would appreciate if you can share us how to interpret the results. Thank you.
Rika Sandra ....greetings please send me an email to discuss further
Hi :) I have exactly the same problem. My independet and my moderator variable ar both dichotomous. My dependet variable is metric. Is this working like in the video? Thx in advance!!
Statistik Auswertung hi please email me personally to discuss this
Thank you for your video! It helps me a lot in doing my theses. I have a few questions to ask:
1. Is the p-value of the interaction below 0.05 to be considered significant? My 'coeff' in the interaction column shows '-.9701' and the p-value of the interaction is '.0048'. Is this considered significant?
2. I'd watched your video about the continuous moderator analysis as well. My second moderator is the age of the participants. I remain it as a continuous variable and run the analysis. The 'coeff' in the interaction column shows '.0928' and the p-value for the interaction is '.000'. Is this considered significant? If it is significant, how should I generate the graph? I've used the same method you taught in this video, but the spss doesn't allow me to set the marker by using a continuous data. Should I group my data and make it become a categorical variable?
3. Are there any assumptions that need to be tested before carrying out this two analysis?
Thank you so much. Hope to hear from you soon.
1. yes it is significant -
2. Yes it is significant.... try to set the markers in the data to draw your graph to see the changes it would be easier that way
Cheers and all the best
Thank you for your kind reply. So I assume that I don't need to test out any assumptions before carrying out the analysis?
yes
How can i get process on spss? In my spss version there is no such option
Hi you need to download from www.processmacro.org/index.html look into the mediation video on this list for details
@@derekong1609 thank u
Hi, is the same process for mediation? coding 0 and 1 and running model 4?
Dear Derek,
Further to Sophie Ho's question on age as scale variable, we have categorised age as millennial, Gen Y and Baby boomers and our respondents have selected their respective category. In order to apply this variable for moderator effect (with 3 categories) what steps should I follow? would appreciate if you could guide us through.
Best regards
Jyothika Shetty
Unfortunately I have not been able to venture into more than 2 categories.. so instead I create an interaction variable for moderator and then run a linear regression to see if the interaction is significant ( if so then there is moderation effect)
HI Dr.Ong. Thank you for the video, it's quite helpful. Could I get some direction from yourself on the following scenario:
I'm planning to conduct a study that'll have 2 categorical independent* variables with 2 levels each, 1 moderator variable , and 1 continuous dependent variable. The results of my moderator (which will most likely be measured on a scale of 1 to 7) are to be scored as low (1-2.9), medium (3-5), and high (5.1-7).
Is it possible to use the PROCESS macro in this case wherein my moderator will have 3 'levels' i.e. low, medium, and high? How should I assign them values if there are 3 'levels' with PROCESS? Thank you in advance.
Note: I am still undecided on whether I will propose my variables as being correlated or as predictor/criterion variables.
Greetings I have yet to explore multicategorical moderators with PROCESS at this stage
I see..thank you anyway.
How is there is no PROCESS Option in My SPSS 25 ?
in SPSS 25 goto Extensions, scroll down to Utilities, then Install Custom Dialog (Compatibility model). Locate 'process.spd' in the Custom dialog builder folder. Select it, it will ask if you want to Install, and, of course you do. Let me know.
Hii Derek, thank you for the video, it's extremely helpful!
But- I'm currently using SPSS version 23, in the Analyze > Regression, there's NO option for PROCESS.
What can I do??? Is there an alternative way to measure the independent, dependent and gender as you've presented in this video???
Thank you!
Evauk88 greetings please check my video on mediation analysis and how to install the extention of Hayes Process
THANK YOU DEREK!! And thank you for the speedy reply. Got PROCESS :-)
Please could you also advice,
would the outcome be violated if I run regression with non-parametric data?
If so, what other ways could I do apart from removing outliers and adjusting outliers?
(I have already tried these two options)
Thank you!!
What if I have 2 controlling variables here? Do I have to report them too?
If you have controlling variables it would be good to report them...unless they are already controlled in the regression
Hello,
Great video- thank you for putting this together. One quick question: Does PROCESS automatically centralize your continuous IV, or do you still need to do that before conducting the moderated regression?
Guy Downs PROCESS automatically center the data for you, i found this from Hayes's book.
Hi @Dr.Ong can I do an moderation analysis if my outcome variable is on a scale of 1-10?
Hi first of all do you want to treat them.as scale or a categorical variable?
Was hoping to treat it as a scale - 1 to 8 represents students grades.is this ok?
@@tyriqueking8873 you may not have enough variance in between if your data is in whole numbers
Derek Ong is there a way to treat it as a categorical variable?
@@tyriqueking8873 treating it as categorical variable would mean to treat it as multicategorical .. for that I am not entirely sure how to analyse it in Process
Hello Derek! Thanks for this great presentation! I would like to ask you if it is possible to share the paper used in this example. Which scale dou you use in order to measure "Perceived Value" and "Product Design"?
Once more, thank you!
Hi this research was not published cheeers
hello Derek and thank you for this demonstration, but does this methode work if the moderator have more than 2 categories? would you please answer me because i need it ASAP
Yes it does although the calculation would be a little bit more complicated.. I havent had the time to fully engage on its complexities so i would like to leave you with this paper which is very useful and does a simulation of multicategorical moderation
Andrew F. Hayes & Amanda K. Montoya
Pages 1-30 | Published online: 18 Jan 2017
A Tutorial on Testing, Visualizing, and Probing an Interaction Involving a Multicategorical Variable in Linear Regression Analysis
Communications Methods and Measures Journal, 11(1), pp.1-30
Taylor and Francis
doi.org/10.1080/19312458.2016.1271116
www.tandfonline.com/doi/full/10.1080/19312458.2016.1271116?src=recsys
All the best
Derek Ong thank you very much for your fast reply, but I have problems at it's application it says error even when I used it with gender (i used 0, 1 codation)
if there are more than 2 cats use 1, 2 n 3 coding
actually I did
oh i'm so sorry Derek, i'm using the latest version of PROCESS v3.0, and at the place of the moderator variable here in your version there is at mine mediator var..that's why it's giving me an Error.. but still, should I put my moderator at the moderator window W or Z below because I've got really confused!!!
hi sir. i first want to be thankful to you for this video as it helped me in doing my thesis. i have a question that i have used gender as moderator and beta values are negative for gender and for interaction , both are significant . i want to ask how to interpret these negative beta values for gender and interaction.
Nida Raja greetings it just basically means that it depends what you coded 0 and 1 as.....meaning that the relationship between your gender and dependant variable is stronger for the lower number than with the higher number...
thank you so much sir i got it.
Bro..what if the beta of the dichomotous variable alone is positive but interaction term is positive. Assume both beta are significant
it can happen when both the effects cancels out each other ..just interpret what the beta tells you cheers
How if i have 5 categories?
At this stage, Hayes has not showed how to do this online with multicategorical moderator. However he has mentioned that the new version and tutorials will be able to do this.. unless you would like to explore the possibility of using dummy variables for comparisons you want to review this paper www.tandfonline.com/doi/full/10.1080/19312458.2016.1271116. Hope this helps cheers
Okay! Can the conditional interaction be significant for all of the categories but the p values above are non significant?! Does it look logic to you?! Not 1, not 2 all if them!!
My apologies. I do not understand your question. Perhaps you may want to send me an email and we can take this offline. Cheers
again Derek, I'm so sorry it worked for all the variables but when it comes to gender it says it's dichotomous and this procedure can't be used
if you don;t mind... you may send me your data ( only send me your independent, dependant and moderator variable and let me try it for you to see what is the problem
Derek Ong good morning Derek..thank you very much, it worked finally, it seems that after doing the reg for other multicats var the model had changed and I wasn't paying attention to it, once I choosed model 1 it worked! It was a crazy day yesterday
Aicha Bensaad okay all the best then