Moderated Mediation and Mediated Moderation: Using PROCESS

[Music] [Applause] good morning good evening depending on [Music] my timeline who are in u.s. or in some other countries so whatever time zone you are in good morning good afternoon and good evening my name is Ryan Ali I'm an assistant professor and graduate program coordinator at College of hospitality and tourism leadership at University of South Florida Sarasota Manatee most of you already know that we are doing these webinars and we invite different scholars from around the world to talk about their expertise and these are all free of course just so that a wider audience can benefit from these experts I recently started a YouTube channel which was like three months ago I named it research base I've got mixed reaction on this name but I decided to stick with this name so far we have 1900 past subscribers I've uploaded 19 videos on on the channel which involves average include different tutorials motivational videos talks about British methods and publication tips and tricks so if you have not subscribed yet to the channel I would highly recommend you subscribing to the channel today's speaker I I recently did some videos on smart BLS and pls so I got some questions about mediation and moderation and then some people also ask me about mediated motivation of moderated mediation so I was looking and scouting in my friends list for somebody who did this and then I dr. Francis Troy so I know him for quite some a few years now we met in Malaysia at one of the workshops and he has done amazing stuff since then so currently dr. Francis is a senior lecturer Abdullah Graduate School of Business at University of Cairo Malaysia he holds a PhD in management and is actively involved in paper from the and I look up to him for some of the stuff that he do especially with very limited resources that he has he's doing a wonderful job in Malaysia he's also acting as the COO managing editor for The Journal of Applied structural equation modeling which is an interesting journal because he also sits on editorial boards for numerous journals and he is quite good at application of pls and also his expertise is in moderation and mediation using process model as well as other methods of mediation and moderation so that's the reason why I invited him today to talk about motivation in mediation and he will also be kind to show a spectacle II how to do these methods and how to interpret the results of mediation and moderation so dr. Francis thank you again for accepting my invitation and I'm going to stop sharing thank you all right thank you dr. faizon is it the my voice loud and clear yeah it's just all perfect thank you all right okay so I will be sharing my screens here trying to get all of this I'm basically looking at my screen right yes per second all right okay so thank you dr. faizon for inviting me for this webinar basically what what is being asked for me to present a is on moderator and mediate the moderation and there has been quite an argument discussing on on which is which right and and probably they might require some clarification on on on all this right so again thank you dr. ferzaan for introducing me myself is Francis coming from osmani of Graduate School of Business University of Durham Lycia right and - together is my email all right okay let's have a look aren't moderation and mediation right before even we start talking about moderator mediation or mediated moderation okay basically what happens here is that in the context of a quantitative research right normally we are trying to examine the association or relationship between two variables namely the independent variables and the dependent rival right which is the core foundations of what are we trying to examine given the inconsistencies of this relationship between the predictors and the outcome variable right we try to introduce a possible mechanism that would have explained this inconsistency right one of it is that probably the relationship between the independent variable and dependent variable at this store right or it's not as straightforward and it is that means you said in between there might be certain mechanism that works within the relationships so so that - relationship between the IV's and v's will be in consistence and given today what we have known from the literature is that well when you introduce a mediator it can be a complementary mediator or it can be a competing mediator depending on the functions of the designs of the indirect effect versus design of the very effect the other way around of looking to the context of relationship between two variables is that we are looking at a possible relationship that this relationship between the two variables is contingent upon a third variable that is outside of the context of the relationship and this we refer this to as external factor let me so said we are postulating that possibility that there is a moderator that means even at different values of these external factors the relationship between the independent variable and the dependent variable varies all right so basically when we talks about a moderation mediation this would be the kind of possible simple models that we are trying to examine whereby we are saying that there is a mediators that acts as a mechanism to breach the relationship between the IV's and v's or that this relationship within the contacts is contingent upon an external factor cause the moderator right next up what I'm going to present or I'm going to share with you today on moderated mediation basically is based on my readings and my understandings from the work of Professor Andrew f his rights which talk about mediation moderation and when he talks about moderated mediation or mediated moderation basically a professor has refer this as the conditional process analysis right so moderator mediation and mediated moderation two terms that I think dr. ferzaan you have in reverse and these are always interchangeable and people always get confused of what actually does all this means all right I will start with what we call it as a moderated mediation basically a moderated mediation means that a mediation that is moderated quite confusing right moderated mediation is a mediation that is moderated or in another word we would have said it that basically a moderated mediation is a moderator indirect effect right this is the by grammatical representations of what is a moderated mediation it's alright assume that we have a mediation model whereby we have one IV one mediator and one dependent variable alright in this case what happens most of the time is that we will say that this is a mediation model that we test for a mediation and then in between the relationship of the IV's and the mediator or the mediator and the dependent variable we put it in a moderator right and if I were to demonstrate or able to show cast this model most of the people will come and and say okay I know the hypothesis the first hypothesis will be the mediator mediates relationship between IV and Evi the second hypothesis will be possibility that the moderator moderates the relationship between IV and the mediator and the third hypothesis will be the media are the moderator moderates the relationship between a mediator and the dependent Bravo and the analysis will be done as such however if you are looking forward towards a moderated mediation what actually this means is that this particular sections is a mediation and a moderated mediation means that this mediation is contingent upon these levels of moderator that means to say we are trying to invest whether the strength or the indirect effect of this relationship differs along different levels of moderator right so this is what a moderated mediation means next up mediated moderation right conceptually a mediated moderation means that a moderation that is carried through mediation right even the same examples that I have demonstrate earlier whereby we have one independent variable one mediator and one dependent Bravo the models remains the same that the relationship between the independent variable and the mediator is contingent upon a moderator all right what happens in the mediator moderation is that we are trying to investigate a moderation that is carried through a mediation that means to set our independent reiber now is the product term of the independent variable times the moderator that effects the mediator and then lead to the dependent Bravo and according to Professor Hays he would have said that a mediator moderation is neither interesting nor meaningful because the product term of the independent variable and the moderator is meaningless let me say there is no point for someone to really test a mediated moderation because the product term of the IB times the moderator itself doesn't represent anything so to his understanding and n22 to what I had known so far is that there is only one model on on this kind of test that that is to say for me it will always be a moderator mediation there there is no possibility of we talking about mediated moderation because according to this equation this relationship doesn't make sense right so why do we examine the director of something that is meaningless this was clearly highlighted by Professor haze in the book in pitch 387 so regardless of whether your model is on the first stage or on the second stage of analysis we are always examining moderated mediation okay so given this model earlier right we consider this as what we call as the first stage of moderated mediation right that means who said that in the statistical equation we are saying that the slang of the indirect effect the mediation effect of this relationship is contingent upon the level of moderator when it interacts in the relationship between the IV and the mediator this is what we know as a first stage of the moderation right as for the second stage of a moderated mediation the moderator is said to have influenced the strength of the mediation at the second stage whereby the moderator moderates the relationship between the mediator and the dependent rebel right okay and when if I sign you have started to propose the webinar session of today's session somebody was talking about the index of the moderator mediation right the index of moderated mediation basically refers to the slopes of the moderator mediation bear in mind that regardless of whether you are doing a second stage moderated mediation or a first stage moderated mediation what we are trying to calculate or we are trying to examine is to look at a linear function of an indirect effect on different levels of moderator right so it is still regression simple linear regression line so the index of moderated mediation refers to be slopes of the regression function all right and for everyone who is interested to work on moderated mediation basically professor Hayes have come up with what we call it as the SPSS processed macro in which in that particular software they have developed a possible of 74 models for you to conduct a moderated mediation and and and later on I will demonstrate to you how a particular moderated mediation can be done as seen in these slides of mine you would have seen that I basically listed up the steps to conducting a moderated mediation first you need to identify that the research frameworks that you are trying to do and match it with the syntax that has been developed in a process macro so if you are doing a first stage moderated mediation it will be a model 7 according to the process macro right and how do you actually get the index of the moderated mediation is by calculating the production of a three times b1 that means to say the product terms of the interaction effect with the past coefficient of the mediator on the dependent variable right and in order to conclude whether a moderated mediation is significant you will have to refer to the upper bound and lower bound of the moderated mediation index right so this is the first stage if you are doing a first stage moderated mediation it will be a model 7 and if you are doing so and if you are doing a second stage moderated mediation it will be model 14 the one that I'm presenting here is a simple moderated mediation model of course you would have expect that you might have multiple mediators or multiple moderators on the same path coefficient that would have in very much differ depends on your model specification all right so that if some brief explanation that I'm able to give on on moderated mediation and to the best of my knowledge I would say that I would have a glib it what prophase had to say that a mediated moderation doesn't really make sense right and I would like to also share that well in order for you to really see so understand moderated mediation you might need to work more on understanding the syntax right that means to set the regression functions by by by by being able to understand the formulas as well as the equation you will be able to understand what what does it mean by a moderated mediation however as being cautious by professor haze right you can do moderated mediation as long as you have deeper the most important part is that when you are hypothesizing a moderated mediation hypothesis you must have a strong theory to say or to explain that why that particular mediator relationship or why that particular indirect effect is contingent upon the levels of moderator right I think the next step will be that I'm going to try to share one of the examples is it on screen no divisor yes is it showing my SPSS yes yes it is but before we do that just a quick thing yeah there are a couple of don't want them now all you want to leave it for the end okay we can try to answer if I'm able to okay let me tell you I mean these questions may be very simple and then may not even need an answer but still I mean some people put them here yeah sure sure first question is can a mediator be a moderator can immediately be a moderator yes well I would have seen yes mm-hmm I would have said yes but it cannot be a mediator and a moderator in the same framework a construct itself can be a moderator on one framework and a mediator on a different framework because it is still control if you posit it as a moderator that means you said you are more interested to to understand that what happens to a relationship at different level of the constructs but if you are positing that variable as a mediator then you are simply saying that this variable serve as a mechanism to explain a distal relationship between two but it is quite illogical for you to have a mediator that serves as a moderator at the same time in a single framework it's really true I actually I agree to you because you know now I visited a few universities and I'm not going to say which University I'm talking about but this entire university and I realized that many students asked me about the research models and stuff and I did them but then after a week or so they in their moderate with their models adding a moderator in it and I was asking them why did you add this moderator in it then they said because my supervisor said it's making your model look better so you know many people still go with adding moderators and mediators not because of the logic behind it or not because of the need for it in terms of research gap or whatever but more because it makes model look more complicated right now the second question was if my moderator is not significant is there any other analysis or description I should provide for it okay if the moderator is insignificant there could be only one possibilities right the only possibilities that that a person would have not a significant moderation probably is that the data range that he has for his moderator is not conclusive that's what a queen is had mentioned if you really wanted to test a variety you need to get a response that covers a range of skills for example right you are saying that you want to test a rival as moderator that means you said in your sample response it is best for you to have responded that answer one and and four five or seven mm-hmm if let's take majority of your responses screwed excuse or askew let me say well you have a region of one to five or one to seven but most of the respondents is answer three four five or five six seven you do not have what we call responses that is heterogeneous homogeneous homogeneous I suppose also the the statistical analysis would not be able to differentiate between high and low and the calculator baby equation and so so what you can do when your moderator is in significant is that you need to relook back into your standard deviation of your moderator as well as the minimum and maximum interval of your response for your moderator this is what ever exceed okay thank you so again there are a couple of other questions and then one is without mediator how many moderators can I have in my model this one I don't know okay how many moderator as as many as you can provided that you have enough samples and your your moderator is telling you you have sufficient power analysis to conduct the moderation the most important power analysis okay okay so now there is two other questions one of them is this person is saying it's a tool using my my question is that I applied a model in BLS and I stood latent variable scores should I go now with these scores in process or should I work the same in smart pls itself for mediation and motivation effects okay it's a very interesting question depending on your hypothesis if you are just saying there your hypothesis is basically that moderator moderates either young mediator and your DB or your IV to a mediator then there's no need for you to move into process macro because my pls will be able to do the job mmm if you hypothesizing saying that that particular mediated relationship in the ten model is contingent upon the moderator then you would have used the little rubbish course and do it in what we call it the process macro but the issue will be whether the results are reliable because I have a concern and and and I'm actually writing a paper lot of eyes on on this because remember when we use my pls we might have a few IDs all right and but when you do process micro you only enter one IV at a time mm-hmm so how reliable will be the mediator moderator medicine mm-hmm so what I'm doing now is basically I'm working on the paper to to allow us to manually calculate the index or moderated mediation when we have our results from an SEM model because whatever whatever that is being done in process macro you can only enter one IV at a time correct all right so by having certain equations and using some calculation now basically we can estimate a moderated mediation relationship in an SEM model that comprises of multiples IVs cool so when do you think you are going to bring this paper out I am going to write this and I'm going to prepare probably I've got to present these people in in Saheb coming August and then I will try to talk without clicks Marku and and Christian to see what they think and and it is possible that we can put this directly into smoke pls I don't know but I have I've been using the the the equation that I derived from basically these paper right so I I managed to do some manual calculation for a complex model okay perfect so good luck for that paper dr. Francis so last question which came here on Facebook and then we can go to the practical one this question is I also put on the route Iseman for this webinar he say that how to report the moderation moderated mediation relation when index is negative and bounce both are on positive side of zero when index is negative and upper bound and the lower bound are both positive that means to say that particular media mediating relationship is contingent upon level moderator hmm but as the functions of the linear regression that relationship would be negative right that means you said the mediation getting stronger when the level of moderator is getting lower mm-hmm right so that makes it the mode the mediation is depending on the level of moderator okay perfect thank you dr. Francis for the answers so I think we can go ahead and do the practical example one time very quickly okay all right yes all right so I will be demonstrating probably something like this the second stage of the moderated mediation model which by using the syntax given in SPSS macro it will be based on model 14 so what happens here is that I have a sets of it right which I have basically taken in from the smart pls program so we are basically let's give a simple model basically we are hypothesizing that Co MP goes to customer satisfaction and goes to customer loyalty and that immediate relationship is depending on switching costs so switching cost is moderating the relationship between customer satisfaction and customer loyalty so what you will be now is that we will be using the SPSS macro that is installed under the regression dialog so this will be the interface of the process macro so your X variable will always be your independent rebel right you're dependent rebel will be Y and then we just plug in whatever that is necessary and then we have our moderator W and moderator Z what happens is that you need to refer to the syntax here if you have a first stitch moderation then you need to plug in your moderator as W if you have a cell for one moment just to go back on what you were saying earlier about your paper so this is that interface and we you were saying that you can only add one independent this is just for people who probably didn't get it and yeah alright so this process macro unless you are able to do with the SPSS syntax then you can probably many play but if you're using a dialog then yes you are only using one IV at a time right so what happened is I now specified this the moderator under Z okay and then I would set my model into model 14 the model number 14 with 95 percent completed interval as well as 5000 bootstrapping right and then we'll just run it [Music] okay I think I may let me see probably see let's see let's try again okay analyze regression but if I were to put it here okay now we call it alright so what happened is that this will be the outcome of the what we call it the moderator mediation or I prefer to use it the in the conditional indirect effect right so this is model 14 which reflects that your model is something like this I think we cannot see the output yeah I think you have to show you okay this one oh yes now it's over yeah all right so we have a model 14 which basically is that your model is something like this okay you'll meet your moderator is moderating the relationship between the mediator and the dependent variable so what the process macro result would give you is that first it will give you the result between the independent variable and the mediator right and the relationship is significant so competencies is my IV customer satisfaction is my mediator so you have the first relationship here the second outcome variable will be looking at customer loyalty which is your ultimate dependent variable alongside with the effect of competencies as well as customer satisfaction and switching cost as well as the interaction effect so you can interpret your interaction if they are moderating effect by looking at these results which says that you have an interval of negative and positive that means to say your moderating effect is non significant all right next up it would tells you the ask questions if you have a moderating effect then you're asked where change you have some ask a change since I don't have any moderating in fact then my ask questions is zero then the next thing is that it gives you a reading of the direct and indirect effect of X on Y right the director of X on Y is equivalent to this let me so said what is the direct relationship between x and y or C Prime in the equation y a then following true the conditioner in the effect is basically telling you what is the mediating effect is the mediating effect or is the indirect effect significant or insignificant at different levels of moderator so in this case we can see that at low level of switching cause the mediation is significant because both are positives the confidence interval at medium level of switching course it is still significant they still mediating and at high level of what we call it the moderator we still have a mediation effect so what this implies is that the mediation effects doesn't changes in line with the levels of moderator so we would have here is that we have the index of moderated mediation which is the slope is close to zero and that we have the level of bootstrap confidence interval they have a negative sign on the lower lower bound and positive sign on the upper bound implying that the mediation is not contingent upon the level of moderate so if a moderated mediation is significant what would have happened is that you have the bootstrap confidence in the load that can be either positive on both sides or can be negative on both and your index of moderation can be either positive and negative and you will start to realize that at different levels of moderator the indirect effect changes so sometimes let's say you have a significant moderate mediation it might be that the mediation is not significant a low level of moderator but a significant at medium and high level of moderator then that by that you can conclude that the entire mediator relationship is depending on the level of moderator all right yeah all right so that is basically how we interpret the BBB results or the output of the SPSS process macro and how we interpret the index of moderated mediation yes perfect okay so now just one other thing to ask so this is one of the limitations as you can do this only for one independent and if your model has multiple independent variables then you have to do it separately for each of them yes but but I wouldn't believe that the result will not be contaminated as compared to you are taking a little rubber score and then do we it process macro one by one I would have advise whoever wanted to do a moderated mediation using multiple IV to to take the results directly from the PLS or SEM analysis and do the manual calculation with the equation given for example that's why I always say that it is important for you to understand this equation you know because these are past coefficients by using this you are able to calculate a moderated mediation for a complex model that has multiple IVs as well as we did all right okay so again dr. Frances I'm really grateful for you to take your time and explain this stuff obviously these are just the introductory concepts and your people have to practice this to go further and you know understand this more and more just one thing in the end I want to ask you and that is IOT if anybody has a question and they email you about that question yeah yeah try to answer whatever that that I am able to okay perfect another thing is many people ask about the PowerPoint slides so normally what I tell them is that it's the speaker's ability but I think if they want can they take some of the things from your PowerPoint slides with reference to you and him and the original slides well I am not quite sure about that that means you said are they asking me for the baby slice or what normally people ask and what I tell them is if the speaker allow it's okay if not obviously it's because the ability but if let's say they are watching the video and they want to take something from the slides watching the video not like your PPT slides but if you're watching the video with reference to your name is that okay yes yeah that would be fine I think the video would have been able to help yeah yeah so it's always good to ask the speakers because you know this referencing and these things are not very common in some parts of the world so people just take whatever they find but I think important to give proper credit to the person who made the slides and so all right then thank you very much again dr. Francis and I hope in the coming months I didn't write you one more time to talk a bit more in detail about these things okay all right thank you [Music]

Channel: Research Beast

Views: 9,201

Rating: 4.89011 out of 5

Keywords: Research Methods, Research Methodology, PROCESS, SPSS, Mediation, Moderation, Moderated Mediation, Mediated Moderation

Id: X4F0zzEhIv4

Length: 43min 47sec (2627 seconds)

Published: Sun Jun 30 2019