SiSSythreesome [build 20080520, experimental]

to download this tool for offline use, simply save this page as "complete website" (or similarily called).
click here to close this
!!! the tool is still experimental and has not been tested extensively. therefore please use it with utmost scrutiny and cross-check the results by median/tertiary split approaches (which, of course, are otherwise evil and should under no circumstances be taken to be endorsed by me). please write an email to johann if you find errors and bugs.

notes and disclaimer

click on the notes to close the notes
the SiSSy threesome (Simple Slopes Syntax for SPSS for three predictor variables to decompose some three-way interaction) tool will generate syntax for spss version > 12 [or below that if you restrict variable name length to 8 characters] to test a threeway interaction and resolve it into 'simple two-way interactions' and the associated simple slopes.
NOTE: this tool can only deal with continuous and dichotomous variables. categorical variables with more than two levels cannot be handled (even though such a version is planned - please check back at occasionally).
NOTE: please be aware that the code looks messy as it will create 30 new variables in your data set and perform 9 individual regression analyses. have spss show the syntax in the output as the comments in the syntax are vital in finding your way around in the results.
NOTE: the code will yield simple slopes for predictor 1 (i.e. relationships of predictor 1 to the dv at various values of predictor 2 and the moderator). please match your variables accordingly with "predictor 1", "predictor 2", and "moderator" in the input here.
NOTE: the predictor and moderator variables must largely fulfill conditions of multivariate normality [and sizes of groups from dichotomous variables must be approximately equal]. if these assumptions are substantially violated, the results of the analysis may not be worth a damn.
indicate your variables and their standard deviations here
all of the following variables must be at least centered if they are continuous or coded [-1,+1] if they are dichotomous!
predictor 1:
predictor 2: -- dichotomous? [check if yes] sd of predictor 2: [enter 1 if predictor 2 is standardized]
moderator candidate: -- dichotomous? [check if yes] sd of moderator: [enter 1 if moderator is standardized]
dependent variable:

if this does not disappear after at most a few seconds, you probably have turned off javascript.
please allow the execution of javascript, the tool needs it to run properly.

please note!!!! this tool DOES NOT WORK IN INTERNET EXPLORER!!!
it technically works like it should (with no guarantee that this is conceptually right) in regular web browsers like mozilla firefox 2, opera 9, and safari 3. it also looks approximately like intended in the latter two. as usual, however, microsoft internet explorer (yes, that's version 7) does not properly execute the javascript needed. so don't use ie, this tool does not work in it. use firefox: