hi everyone my name is amanda montoya and i'm an assistant professor of quantitative psychology at ucla and hopefully you are watching this video because you are a beta tester for memory version 3. and i decided to make a series of videos that go along with this beta release mostly because i learned people don't read documentation right so um i thought maybe a um video version of some of the kind of key points in the documentation might be a little bit more fun and interesting way to learn about memory so um in this series of videos there are a variety of videos this the first one with a general introduction to memory and some changes to version 3. there are then a few different videos on how to install memory depending on which way you are using it in spss sas or the dialog box i'll walk through the model templates and how those work um the model templates have been around since version two but they are wildly expanded for this version um i'm then going to talk specifically about moderated mediation models which is the big change with this new version of memory where you can now estimate moderated mediation models and then um the second to last video i uh walk through some of the nitty-gritty of how the plots of conditional effects work just because the details of that are not um documented anywhere and so i mean they're in the documentation but um there's no papers that really talk about how to do this or why we do it the way that we do um so i wanted that to be out there and then i will wrap up with a video with just a couple um details if some of these things are completely not of interest to you you can skip the videos that's why i've made them kind of separate videos but the idea is that this is kind of a whole playlist since apparently everybody's a youtuber now um so let's get started uh so what is memory memory is a macro for spss and sas the name comes from mediation and moderation for repeated measures designs um hence the name memory i'm sure many of you out there are having a um a moment where you're like oh my gosh i've been saying this wrong for so long i don't care how people say it as long as they use it so um but the way that i say it is memory and um it's a macro for spss and sas that then allows people to do mediation moderation analyses in repeated measures designs that aren't necessarily possible with the built-in functions um in spss or sas or at least aren't very easy with those built-in functions so uh these can be used for models with continuous mediators and continuous outcome variables at this point um i'm not doing anything with dichotomous outcomes though i do have a student who is working on this um and if you're interested you're always welcome to reach out to me we would love some practice data and what this model does is it uses different scores to model the impact of a repeated measures factor on an outcome so in memory the kind of causal variable of interest is always the repeated measures factor that differentiates the repeated measurements so um if you are in a case where you're actually measuring your uh x variable uh then memory is not going to be appropriate and you can skip all the way to the wrap up video where i will point you to some other resources that are helpful if you have um a different kind of x variable so um what's new in this new version of memory well so this is version three version one included just mediation versus version two added moderation so a surprise surprise version three includes moderated mediation and i'm really excited about this this has been a culmination of a lot of work um so there are now models four through 18 in memory and um the model numbering system is is kind of modeled after um process which is a tool that lots of folks are familiar with um for for doing mediation moderation analysis in between subjects designs but basically what we've done is we've added a number of models that can incorporate one moderator into a simple or parallel mediation model and it is that's the big new thing um and then an additional thing that i added this time is a little bit smarter mean centering so now with the center option the default is zero which means that none of the moderators are centered center equals one will center all of the moderators including dichotomous moderators which i think is kind of a funky thing to do um so i added a new option which is center equals two uh which would center non-dichotomous moderators so if you have some moderators that are continuous and some that are uh dichotomous um send the center equal to option will uh center just the non-dichotomous moderators all right and um a couple things have changed in this new version so i have actually removed the bias corrected bootstrap confidence interval research overwhelmingly shows that this method has inflated type one error rates meaning that it's giving false positives and given the kind of state of the world in psychology and behavioral science research it just didn't seem right to me to include an option that we know performs poorly in this manner so the bias corrected bootstrap confidence interval is gone um but the percentile bootstrap confidence interval and the monte carlo confidence interval are still available um the percentile boost drop confidence interval is the default and both of those methods have been shown to have much better and more accurate type one error rates so um so if you're used to using the vice corrected you can just switch the default the other thing is that in uh doing these um changes we discovered an error um it's kind of amazing to me that we only found one but that i guess speaks to the testing process um if you uh used normal standard error estimates so if you use the normal equals one option with a serial mediation model um the standard error calculation may have been incorrect um i've only found one person for whom this affected but if you have any questions about this you're more than welcome to reach out to me we've double checked everything now and all of these standard error calculations are correct for the serial mediation models this did not affect any of the parallel mediation models um and it only affected folks who are using the normal option um which is also not really recommended so um so yeah so those are two changes and that's it for this intro video so