Видео

Everything I report in the video is crucial for psychological research, though it may differ in your specific field of study.

Reporting would not hurt

no. [you just need to send your second moderator to the "moderator variable Z" box.]

Knowing the total effect can be useful. it is the sum of direct and indirect effects. it would not hurt to report it.

Not yet

There is no definitive answer to this question, since I do not know your data and results, but the highest order unconditional interaction analysis tests the simultaneous null hypothesis that the regression coefficients for all of your interaction terms are equal to zero. The p-value for each of the terms in the model section deals with them separately. As I said, I cannot answer this question definitively, but I would say that you still have a significant interaction that can be interpreted.

Good question. Bootstrapping is a resampling technique that involves drawing random samples from the original sample. Each time we bootstrap, we may end up with slightly different confidence intervals due to the random nature of the resampling process. It appears that my write-up in the Word file was based on a previous run of the analysis. In the output, shown in the video the interval is .049-.201, which is very close to the 0.049 to 0.197 range you can see in the Word file.

@@AppliedStats Thank you so much! I have another question, from my analysis my c' path p value is 0.913. (path a and b's p value are .019 and < .001 respectively). In this case, can I interpret my model is significant? (indirect effect is significant though)

if the ci for indirect effect does not include zero, you have a significant mediation

How about watching my new video on binary moderators: ruclips.net/video/AIKiV30kX30/видео.htmlsi=1NiqdwVfvJuU1esI

how about using these keywords? scholar.google.com/scholar?as_ylo=2020&q=%22moderation%22+%2B+%22process+macro%22+%2B+spss+&hl=en&as_sdt=0,5

this means your indirect effect is not significant

@@AppliedStats Yes, thanks.

In order to obtain results even when the interaction is not significant, select the "Always" option located under "Probe Interactions" in the option window.

Ys i can't get conditional effect of the focal predictor at values table...pls help me sir

@@naseerajasmir1234 In order to obtain results even when the interaction is not significant, select the "Always" option located under "Probe Interactions" in the option window.

B = coeff

If your outcome variable or mediator is binary (such as yes/no or success/failure), Process Macro automatically performs a logistic regression analysis. That's why you see outputs like -2LL, ModelLL, and McFadden's R-squared, which are specific to logistic regression. In this video, I used continuous outcome and mediator variables, for which the macro performs linear regression by default.

ruclips.net/video/-030VJ-jidc/видео.htmlsi=jldZbd7ak1DyzPy4

Just need to send them to the "covariates" box

@@AppliedStats Thank you so much! However, I have created dummy variables and, when I send them to the covariates box, after running the macro, it gives an error saying "one of your model variables exhibits no variation (its a constant)". Do you know what should I do then?

​@@cristinamorillo7259 there is a problem with your dummy variable

In order to obtain results even when the interaction is not significant, select the "Always" option located under "Probe Interactions" in the Options window.

If I understand your question correctly, your statistical hypothesis may be: No mediation: The mediator has no effect, the predictor only directly affects the outcome. Mediation: The mediator explains part of the effect of the predictor on the outcome.

@@AppliedStats Thank you but I mean like with regression: ​ H0: β = 0 H1: β < 0 -> how to do this with a mediation?

H0: ab = 0 H1: ab ≠ 0

@@AppliedStats Thank you very much! But what does "ab" stand for?

From the video, ab = indirect effect which explains whether there is a mediation or not. I hope I’m correct?

PROCESS does not allow dichotomous mediators. I am not aware of a workaround. But I also do not think that ethnicity can be used as a mediator anyways.

@@AppliedStats okay, thank you for the reply!

PROCESS accepts a dichotomous Y variable. If the entered variable for Y is identified as dichotomous, PROCESS will employ logistic regression to model the Y variable. The coefficients in the output are expressed in a log-odds metric. Instead of the typical R-squared, McFadden's R-squared, Cox and Snell's R-squared, and Nagelkerke's R-squared are reported.

@@AppliedStats Thank you for your response. I conducted a mediation analysis in PROCESS with a binary DV, a binary IV, and a continuous mediator. My findings are as follows: 1) The first model (X -> M) was significant. 2) In the second model, the X (IV) was not a significant predictor of the Y (DV) but the M (mediator) was. 3) In the third model, the direct effect of X on Y was not significant. The indirect effect of X on Y was significant. What do you make of it? I think this means that the mediator completely explains the relationship between the IV and the DV. So I have a complete mediation, no?

@@AppliedStats One more thing. The same IV significantly predicts the DV if I run a binary logistic regression. Does it then make sense that the same IV does not predict the same DV in mediation analysis? In PROCESS: 1)The direct effect of X on Y is the path which doesn't include the mediator. 2) The indirect effect of X on Y is the path which includes the mediator. Do I understand it correctly?

If the direct effect is not significant and the indirect effect is significant, you have full or complete mediation.

I explain in the video where to get the direct effect and where to get the indirect effect. the third regression in the output does not give you the direct effect (c'), it gives you the total effect (c).

you can send more than 1 mediator to the mediator box

In JASP, it is not possible to directly perform a 3-step h regression analysis. However, there is an alternative approach that you can use to achieve identical results. Instead of running one h regression, you can run three separate regression analyses with increasing sets of variables. 1- First, run a regression analysis using only the variables in block 1. 2- Then run a second regression analysis using the variables in Block 1 and Block 2. 3- Finally, run a third regression analysis using all the variables in Block 1, Block 2 and Block 3. You can then calculate the r2 changes by subtracting them from each other.

@@user-df2fu2tw7y Thank you very much sir, you confirmed our doubts, thanks! Have a nice Day 🥰🤗