Testing mediation using bootstrapping in SPSS

1 To fully understand the following procedures read these key papersHayes, A. F. (2009). Beyond Baron and Kenny: Statistical mediation analysis in the New Monographs, 76(4), 408-420. doi: , K. J., & Hayes, A. F. (2004). spss and SAS procedures for estimating indirect effects insimple mediation Research Methods, Instruments, and Computers, 36, , P. E., & Bolger, N. (2002). mediation in experimental and nonexperimental studies: Newprocedures and Methods, 7(4), , D. P., Lockwood, C. M., Hoffman, J. M., West, S. G., & Sheets, V. (2002). A comparisonof methods to test mediation and other intervening variable Methods, 7(1), is an interactive PDF- if you are viewing this on a computer connected to the internet: the hypertext links (blueunderlined) should take you to the relevant site.

2 Access the relevant datafile by clicking onthe link- if you are reading this off a hardcopy: Really??? How old I guess you'll need toboot up the old machine to run the analyses anyway so you may as well open this PDF onscreen and access the links (see instructions above) Testing mediation using bootstrapping in SPSSC reated by Natalie J. LoxtonPage 1 The macros used in this tipsheet can be found at the see Hayes's Facebookpage(oh, as if you aren't on ) Testing mediation using bootstrapping in SPSSC reated by Natalie J. LoxtonPage 2 For this example we will use the Sobel MacroDownload theSPSS macroto your computerYou should also download the documentation ( )for additional informationAlso note as stated on the website that another Macro " " will also perform the sameanalysis unless you need to estimate a model with dichotomous , for this exercise we will use this note that this version of the macro does not work in spss ver 18 (unless you have all thepatches).

3 The older version (sobel_SPSS) does work in ver 18 mediation using bootstrapping in SPSSC reated by Natalie J. LoxtonPage 3 Let's test this modelThe dataset can be accessedHERENote the variable names that map onto the spss fileAlso recall the mediation conventions (from Baron and Kenny, 1986)Path a = IV to MedVPath b = MedV to DV, controlling for the IV(when multiplied together these are the INDIRECT effect)Path c = IV to DV (aka the TOTAL effect)Path c' = IV to DV, controlling for the MedV (aka the DIRECT effect)Total effects = Indirect effect + Direct effectTesting mediation using bootstrapping in SPSSC reated by Natalie J. LoxtonPage 41) Run the Macro (Sobel)Macros are very similar to syntax files and are run as such:Open the file via the "Open Syntax" optionMake sure to select the file " "This is the file you downloaded from the Hayes websiteTesting mediation using bootstrapping in SPSSC reated by Natalie J.

4 LoxtonPage 5 Run the macroSelect " "This activates the macro (which runs in the background)You can close the macro file if you wish - this will stay in the background until you either restartSPSS or load a different macroTesting mediation using bootstrapping in SPSSC reated by Natalie J. LoxtonPage 62) Set up your model in another syntax fileIn this step you need to tell the macro which of your variables are the IV, DV and do this you need to create a new syntax file and set up your model using this template:SOBEL Y =DV/X = IV/M = MED/BOOT = is the syntax for the example model:SOBEL Y =JOBSAT/X = POSAFFECT/M = WSE/BOOT = this syntaxNote the "boot = 5000"This is the part of the syntax that tells spss to draw 5000 sample indirect effects (with replacement) using your sample data as the "population" of possible indirect effectsTesting mediation using bootstrapping in SPSSC reated by Natalie J.

5 LoxtonPage 7 The macro will now run to test your modelNote. This can take a little while as this is performing rather complex equationsTo check if the program is still running you will see a little "Running MATRIX" in the bottom rightTesting mediation using bootstrapping in SPSSC reated by Natalie J. LoxtonPage 83. Interpret the outputBy default the initial macro will be recorded in the output file (ignore this and scroll to the bottom ofthe output file) Testing mediation using bootstrapping in SPSSC reated by Natalie J. LoxtonPage 9 Direct and Total Effects outputThis part of the output provides the significance tests of the:1) "c" path (IV to DV, aka the Total effect) - in this example significant p = .00622) "a" path (IV to MedV) - in this example significant p =.00003) "b" path (MedV to DV, controlling for the IV) - in this example significant =.

6 00004) "c' " path (IV to DV, controlling for the MedV, aka the Direct effect) - in this example non-significantp = .5159So, this model has met all the criteria for mediation according to Baron and Kenny (1986) - althoughnote that the IV-DV total effect criteria is no longer necessary (see earlier refs)This coefficients are theUnstandardised B weightsIn the latest macro the bivariate correlations are displayed but to obtain theStandardised Betaweightsfor paths b and c' you would need to run a multiple regression in spss with the IV andMedV as predictors of the DV and report the beta weights) Testing mediation using bootstrapping in SPSSC reated by Natalie J. LoxtonPage 10 Indirect Effect (Sobel)This section provides the test of the significance of the INDIRECT effect (path a X path b) using theSobel testIn this example path a =.

7 7356 and path b = .3048So the indirect effect = .2242 (see #1)And is significant using the Sobel test (see #2) Testing mediation using bootstrapping in SPSSC reated by Natalie J. LoxtonPage 11 Bootstrapped effectFinally, the output provides the bootstrapped confidence intervals (99 and 95 percentiles arecalculated but we only want to look at the 95%)Here we are looking to see if ZERO (0) lies within the interval rangeNote that LL = Lower Limit (or the lower boundary) and UL = Upper Limit (or upper boundary) of theConfidence we are asking whether it is possible (with 95% confidence) that the TRUE indirect effectwould be ZERO (basically, no mediation ).In this case the TRUE indirect effect is 95% likely to range from .1342 to .3232 - the estimated effectis .2246 (lying in between these two values)If (as in this case) zero does not occur between the LL and the UL then we can conclude that theindirect effect is significantAn example of a non-significant indirect effect would be to have a LL an UL of.

8 3232(since zero is a possible value between a negative number and a positive number)For an example of reporting indirect effects using bootstrapping clickhere(note that this exampleused AMOS but the reporting is similar)Those wishing more information on this macro and/or mediation etc are advised to refer to the Hayeswebsite or the Facebook mediation websitePlease note that the creator of this tipsheet is not associated with the developers of the macros andis unable to provide additional advice on the macro or issues related with mediation or mediation using bootstrapping in SPSSC reated by Natalie J. LoxtonPage 12