MEDIATION AND MODERATION · METHOD GUIDE
One-on-one help setting up, testing, probing, and reporting mediation and moderation. You stay the author, and you leave able to explain what the paths do and do not show.
Book a free consultation30 minutes · Bring your output, your plan, or your committee's comments.
WHERE YOU MIGHT BE STUCK
Most projects stall at one of three points.
Start with the one that fits.
Decide the roles of your variables before you choose software or syntax.
Test the indirect effect directly and probe interactions in a way you can explain.
Use causal language that matches your design and report the uncertainty.
DECISIONS THAT SHAPE YOUR ANALYSIS
A mediation model describes a causal chain from predictor to mediator to outcome. Cross-sectional data can be consistent with that chain but cannot establish it. Justify the ordering and confounding assumptions, and word the conclusions to match the design.
The indirect effect is a product of two paths, and its sampling distribution is usually not symmetric. Intervals that allow for that, such as bootstrap confidence intervals, are generally preferred to relying only on separate significance tests of each path.
A significant interaction says the effect varies. Simple slopes and regions of significance show how it varies, at values you can justify.
START ON YOUR OWN
Use these before or after your consultation.
How to run and report a simple mediation model with bootstrap confidence intervals.
Read the guidePlan your sample size before you collect data, including designs with more than one path.
Plan my sampleEstimate indirect effects, interactions, simple slopes, and moderated mediation.
See the workflowInterpret conditional effects without treating nonsignificant main effects as a contradiction.
Read the CaseFormat statistics correctly, with italics, spacing, and decimal places handled for you.
Format my resultsCheck which analysis fits your variables and design.
Choose a testACADEMIC INTEGRITY
Dr. Yel does not write your dissertation or any chapter of it. The decisions, the interpretations, and the prose are yours. His role is to advise, teach, and review.
Nothing is handed over unexplained. We work through each method, assumption, and result until you can defend it on your own.
Programs differ in what outside help they allow. Check your policy and let your chair know you're working with a statistical consultant.
Dr. Yel declines work that would misrepresent authorship or data, such as writing chapters, altering or fabricating data, or anything your program's policy prohibits.
A FEW THINGS YOU MIGHT BE WONDERING
No. PROCESS is a popular tool, but the same models can be estimated in regression or structural equation modeling software with bootstrapping. Use whatever your program or committee expects.
You can test whether the data fit a proposed mediation pattern, but you should not claim the data prove a causal chain. Word your conclusions to match the design and explain the assumptions.
It depends on the size of the paths you expect and on your method. Simulation-based power analysis is often the most defensible approach, and the free power tools are a good place to start.
Report it as it is. A nonsignificant path does not automatically rule out an indirect effect, so look at the interval for the indirect effect and discuss the result carefully.
Analyses are run together in live sessions, with you watching or at the keyboard, so you can explain every step. You keep the syntax or code from every session.
The scope depends on your model, your data, and the support you need. The free consultation is the place to discuss deliverables, pricing, and timing before you decide.
BRING THE MODEL THAT’S HOLDING YOU UP
Meet with Dr. Yel for a free 30-minute consultation about your study, data, or committee feedback.
Book a free consultationCan't find a time that works? Email bookingrequest@dissertationstatshelper.com with a few times you're available and your time zone, and we'll find a time. It's a new address, so if you don't see our reply, please check your spam folder.