in this video I'm gonna walk you through the steps associated with conducting a mediation analysis I'm gonna assume that you've already watched a video relevant to concepts and terms in a mediation because I'm gonna walk through some of this fairly quickly assuming that you have exposure to that material so first I'm gonna walk you through the steps in a generic sense associated with conducting a mediation analysis with one mediator and then I'm gonna go through the example with you step by step actually getting the results for each step so relatively briefly here are the steps first step one estimate the total effect between x and y that is going to be between the independent variable and the dependent variable and I ultimately want to compare this total effect with the direct effect of interest that is going to be estimated in the mediation model this is not the mediation model this is just the total effect being estimated and I can do that easily with a bivariate regression step to estimate the direct effect between X and M the independent variable and the mediator now that is this term over here a X and M I need to estimate that direct effect you could consider this a total effect as well but I usually refer to it as I direct effect and the way you can estimate that is with a bivariate regression again just using the independent variable as a predictor of the mediator this is an easy type of statistical analysis to conduct next step three a and I call it three a because there's a three a and a 3b and each step is essentially conducted at the same time and I'll explain in a second so you need to estimate the direct effect between x and y that is the primary direct effect of interest here the effect of the independent variable on the dependent variable and how can I estimate this well you can estimate it in a multiple regression where you include the independent variable and the mediator as predictors of the dependent variable that is how you're going to get this unstandardized beta weight or standardized beta weight depending on the approach you take step 3 be estimate the direct effect between m and Y so estimate this B which again can be estimated the multiple regression where you have X and M as predictors of Y so really the brunt of the analysis is really just one bivariate regression where you have X predicting M and then you run a multiple regression where you have X and M predicting Y and that is how you can get each one of these three coefficients and the standard errors in order to estimate the indirect effect for statistical significance that's step four and a lot of people use the Sobel test and I'll show you how to do that and that's gonna require a calculator either from a webpage or possibly syntax in spss you can also estimate it through bootstrapping and I'll show you both methods okay so now I'm gonna go through the steps with an example and that example is a study where a researcher hypothesizes that the effect between self oriented perfectionism and positive effect may be mediated by flourishing and so the independent variable here is X which is self oriented perfectionism Y is the dependent variable positive effect and M is the mediator flourishing so what does that look like in the model we have X self oriented perfectionism we have Y positive effect and then we have M flourishing and with these variables and these coefficients that need to be estimated we can conduct a mediation analysis with the simplicity of conducting one bivariate regression and one multiple regression and with those analyses we can estimate the direct effect see that's the primary direct effective interest and the indirect effect a times B now before getting into that bivariate regression and the multiple regression that serve as input into the mediation model you want to estimate the total effect to see if there's even any point in conducting the mediation analysis and that is represented with this basic bivariate regression model or I have self-oriented perfectionism predicting positive effect and that is conducted with a bivariate regression between x and y so let me do that now analyze regression linear and put self I entered perfection into block 1 of 1 and positive effect into the dependent variable and click OK and go straight to the coefficients table and here I've got the unstandardized beta way to 0.1 0 2 and it's statistically significant and the standardized beta weight is equal to 0.14 0 so that is step one conducted I've estimated the total effect and it's statistically significant so I can carry on step two direct effect of X on M and that is self oriented perfectionism predicting flourishing this is a key coefficient required for the estimation of the indirect effect and that can be done very easily in a bivariate regression so analyze regression linear self oriented perfection M stays in the independent variable box take out positive effect and put flourish into the dependent box and click OK scroll down to the last table coefficients here's the unstandardized beta weight 0.197 and it's also statistically significant and the standardized beta weight is point one nine zero what I need though are these two values I need the unstandardized beta weight and I need the unstandardized standard error and I'm going to put those values into this slide here so a equals 0.197 and the standard error is equal to point zero five two I'm gonna need these values in order to conduct the Sobel test that is to test the statistical significance of the indirect effect so a equals 0.197 that's very easy to estimate it's just a bivariate regression now I make mention of it as a direct effect it's not the primary direct effect of interest it's really just a coefficient I need to estimate the indirect effect next step 3a and 3b I need to conduct a multiple regression with self-oriented perfectionism and flourishing as the independent variables and positive effect as the dependent variable and I'm going to be able to estimate B and C I right here B is a direct effect not of interest it's not really a particular interest because I just need it in order to estimate the indirect effect whereas C is the direct effect of primary interest so if this is statistically significant it will support the notion that there's a direct effect of self oriented perfectionism on to positive effect now there's a curved arrow here because this is your multiple regression and in multiple regression the predictive variables are only correlated with each other there's no direct effect between these two variables so let me conduct that multiple regression easy enough analyze regression linear self oriented perfectionism I need to take flourish out of the dependent variable box and put it in the independent variable box and now I need to put positive effect into the dependent variable box and click OK and I scroll down coefficients I've got the last pieces of information I need in order to estimate the statistical significance of the indirect effect point zero one seven for self Ironton perfectionism point zero one seven and the standard air was equal to 0.03 0.03 zero and I also need the flourish unstandardized beta weight and standard error 0.43 zero point four three zero and a standard air out point zero two eight all right so now I've got all the terms required in order to estimate the indirect effect as well as test the indirect effect for statistical significance before I do I'm just gonna point out that self-oriented perfectionism the unstandardized beta a two point zero one seven is not statistically significant P equal point five seven three therefore there is no statistically significant direct effect between self oriented perfectionism and positive effect I'm presuming the indirect effect through flourishing will be statistically significant but I have to test that first so next step for Sobel test or bootstrapping in order to test the indirect effect for statistical significance and it can be as simple as using a web page calculator in order to execute the Sobel test however there are also recommendations to use bootstrapping as a superior test over the Sobel test but that would have to be executed in a program and in spss there is syntax available to test the indirect effect via bootstrapping I'm gonna show you both methods so in order to get the web page calculator just write Sobel test calculator into a search engine and you'll see a top one come up by k-j preacher it's a good one so click on Sobel test calculator you can just scroll down to this first table here and what it's asking for are the unstandardized beta weights associated with the indirect effect a and B and I estimated those a and B right here so that's what it's asking for so I'm going to input 0.197 and be the estimate between flourishing and positive effect 0.4 three zero point four three zero and that's also asking for the standard errors and I have those as well point zero five two point zero five two and point zero two eight point zero two eight then I click on calculate and the Sobel test right here so test is associated with a Z value of three point six eight and a standard error of 0.02 three and a p-value less than point zero five in fact its P equal point zero zero zero two P less than point zero zero one therefore I can conclude that the indirect effect between self oriented perfectionism and positive effect via the intermediary variable of flourishing is statistically significant I haven't actually estimated the point estimate of the indirect effect but it's simply the product of a and B so let me just calculate that right now point one nine seven times 0.4 three zero and that equals point zero eight five that is the estimate of the indirect effect between self oriented perfectionism and positive effect through flourishing and the Sobel test is suggesting that that point estimate of point zero eight five is statistically significant P less than point zero zero one so that's the Sobel test overall that is a mediation analysis with one mediator using the Sobel test to estimate the statistical significance associated with the indirect effect which was estimated at point zero eight five on the basis multiplying the point 197 and the point four three zero now the direct effect between self Orient and perfectionism and positive effect was not found to be statistically significant on the basis of the multiple regression that I conducted recall that that had a p-value of 0.5 seven three the direct effect of point zero one seven this implies that the effect of self oriented perfectionism is only a good thing it only has a potential influence on people's optimism and happiness and well-being if it facilitates them to be productive and feel like they're getting on with their work and doing well and accomplishing things that is how self oriented perfectionism can potentially have an influence on positive effect it does not have a direct effect on positive effect it's only through the feeling of achieving things and being productive now as I mentioned in the book not everyone's in love with the Sobel test if some people argue that bootstrapping should be used in order to estimate the statistical significance of an indirect effect and to some degree I'm open to that argument especially considering that bootstrapping doesn't assume any level of normality so how to do a bootstrap analysis to test the indirect effect for statistical significance you need to get the syntax that I make reference to in the textbook so in the textbook I provide a link and once you have this syntax file open in SPSS you need to include a data file that has only the variables of interest so click new data and you want to copy three variables only into that data file you want the dependent variable put that variable first put the independent variable and put the mediator third so dependent variable which is positive effect so copy and paste and now I want the independent variable which is self oriented perfectionism copy and paste and now I need the hypothesized mediator which is flourish copy and paste once you have the three variables necessary to conduct the bootstrapped approach to evaluating the statistical significance of the indirect effect you need to change the variable names and change them to the following the dependent variable should be called y var the independent variable should be called x var and the mediator should be called M var so with those name changes save your data file once you have the syntax file open you'll want to run it to put it into SPSS as memory so go over to run and all and that has put it into SPSS as memory and now you need one line of code in order to execute the bootstrap analysis to test the indirect effect for statistical significance and that line of code is in the paper from which this program was created so it's called Sobel so just copy that control C and then in the new syntax file click paste and you have to change this boot equals e to a particular value that you want so do you want 1,000 replications mm so let's put 1,000 replications and then run an SPSS has executed the bootstrap analysis and you can see the program first reports the descriptive statistics and the Pearson correlation it even estimates the direct and total effects associated with the analysis much like I did with SPSS importantly it also calculates the indirect effect and the significance now this first row is the Sobel test so this is the Sobel test that I already conducted you can see that the Z value is three point six seven two well that's exactly what I got here three point six seven well not exactly three point six seven eight so very similar within rounding and it's statistically significant P equal point zero zero zero two very similar Oh point zero zero zero two so that's the Sobel test so you don't have to use the web calculator if you want to use the syntax but the real value of the syntax is the bootstrapping so here we've got the point estimate of point zero eight four nine which is what I calculated any calculator within rounding point zero eight four seven point estimate and we can see that the 95% confidence intervals do not intersect with zero the lower bound is point zero three seven five and the upper bound is point one two nine eight and because both of these are positive and the point estimate is positive I would declare that this point estimate is statistically significant and because that point estimate of point zero eight for nine is the indirect effect associated with the mediation analysis I again would conclude that there is statistically significant mediation taking place and because the direct effect between the independent vary and the dependent variable is not significant this is a case of full mediation so those are the steps associated with conducting a mediation analysis with one mediator mediator