Hi,

you should dummy code the IV--that facilitates the interpretation. Further, mean center the SDO variable (for the same reason). With that done, the output will have the following interpretation

1) Intercept/constant: Mean of helping intentions for the 0 group (then: the muslim condition) and average SDO (→ mean centering result)

2) Target: Difference between the muslim vs. non-muslim condition for those respondents with an average SDO

3) SDO: Effect of SDO in the muslim condition (X = 0)

4) Product / Int_1: Change in the mean difference between muslim and non-muslim scenerios with every unit change of SDO (if you standardize SDO its is then an increase in 1 SD).

Better plot the result. Google for the site by Jeremy Dawson who provides some excel sheets for that.

All the best

Holger

Holger Steinmetz, thank you for your response! IV has been dummy-coded (0,1) and SDO is mean-centered, the output for that is the same as the one included in my original question.

Are you interpreting the 'model' table?

For the 3rd interpretation (SDO) - Which value tells you information about the IV here? I thought this row simply tells us whether M (SDO) significantly predicts the DV (bystander helping intentions). Thereby considering both conditions. My interpretation for this regression was: SDO negatively predicted bystander helping intentions; for every one unit increase in SDO there is a -.14 decrease in bystander helping intentions.

Would you be able to explain your interpretation further please? Thanks!

I used the syntax for 'visualizing the conditional effect of the focal predictor' provided in the output to automatically plot the results, which I have attached here.

Based on the output, SDO interacts with the non-Muslim condition and bystander helping intentions to a greater degree than with the Muslim woman condition.

But since the model table indicates for vignette target there are no significant differences between the conditions for respondents with an average SDO - Vignette target is not a predictor of bystander helping intentions (p = .80).

Is my interpretation correct? Also, thanks for recommending the excel sheet for plotting, I will check that out

The significant t-test for the interaction term in your model shows that the slopes of the two lines differ significantly. But at the 3 values of X that are shown in your results (x=-.856, x=0, x=.856), fitted values on the two lines do not differ significantly.

I suspect your output is from Hayes' PROCESS macro, and that -.856 and .856 correspond to the mean ± one SD. Is that right?

Why does it matter if the fitted values on the two lines do not differ significantly at those particular values of X? Your main question is whether the slopes differ significantly, is it not?

Bruce Weaver Yes, It is hayes PROCESS macro, and those values correspond to the mean ± one SD and when the mean is centered for the construction of products.

Yes, you're right it does not matter too much. Since that is the main question is if the slopes were significant.

But just to confirm, am I correct in saying my graph corresponds to the interaction, and the two lines are the two experimental conditions (0, 1)? Just because I have seen in tutorials that when you copy & paste the syntax output for visualizing the conditional effect of the focal predictor, and it generates a graph that corresponds to the simple slopes (see the attached image). But I guess it only presents the data of the slopes (low, average, high) when at least one is significant?

Hi all,

it's getting a bit complex :) I respond step by step (at the risk that Bruce Weaver has already responded :)

You wrote

"For the 3rd interpretation (SDO) - Which value tells you information about the IV here? I thought this row simply tells us whether M (SDO) significantly predicts the DV (bystander helping intentions)."

The relevant parameter is in the first table (B = -.1446). This is the effect of SDO *for the 0-Group". Generally, such a coefficient is called a first-order effect and is defined as the effect of X when Z = 0. Main effects shift into first order effects once a product term is added to the model.

You then presented a slopoe diagram that exactly shows this. The solid line is the first oder effect of SDO of -.1446.

However, this diagram is rather off because your research question is about the group differences and whether they are moderated by SDO and not the effect of SDO.

As I said in my first post: At average SDO (SDO=0), both groups don't differ significantly (B = -.0218, p = .80), however with each unit increase of SDO the difference becomes larger with B = -.2136. However you cannot yet tell wheter one of hte groups results in higher intentiosn or the other results in lower intention. That's what the second table tells you (or an appropriate graph): At 1 SD below the mean of SDO (-.85), the group difference is positive (B = -.16), however non-signficant (p = .1958). That is, the woman condition results in slightly higher intentions, but again: non-sig.

1 SD above the mean of SDO , the difference is turned. Now the muslim-women condition results in higher intentions or stated differencly the woman-condition results in B = -.2046 less. Likewise, this differnce is non-sign. Having a significant interaction while one or both first-order effects is non-signficant seems a bit odd, but the interaction is the angle while the first-order effects is the angle between thes lope and a zero-slope (line that is parallel to the x-axis).

I cannot really understand your second plot, as you seem to use a different variable (support?). Make sure you use the dummy-coded condition variable as the IV (X-axis) and the mean centered SDO variable (and 1SD + and -).

HTH

Holger

Holger Steinmetz oh okay I see now, I was not aware that it was the b coefficient, and that main effects shift to first order effects so thank you for clarifying that! :)

How would recommend reporting the 3rd interpretation in a lab report?

I was thinking something like... "First order effects of SDO indicated that SDO negatively predicted bystander intentions in the 'Muslim woman' condition, b = -.14, t (88) = -2.82 p = .01, confidence intervals between -.25 and -.04."

Then, below the text present the scattorplot, since information about the 'Woman' condition was not mentioned in the text due to the first order effect. It's worth including otherwise the report will be missing information about the 'Woman' group, even though group differences were not significant, no?

Also, the second scatter plot is from a tutorial I watched, I just wanted to confirm that syntax provided in PROCESS output will produce the high/average/low slopes of the moderator (which in that plot is support) only if one of the conditional effects is sig. Which would make sense. I'm new to moderation, and there seem to be no tutorials that use a dichotomous IV. So I just want to be sure of what I am doing/how to correctly interpret things. So thank you for your assistance!

Hi,

it is hard what to recommend. I had the impression that your research focus on SDO was simply that SDO was a moderator. Hence, its own effect would not be the point. If you want to have the SDO effecs of both groups then you can calculate the B's from the table

B_group1 = -.1446 + (-.2136) = -.4272 but you

need to do a simple slope analysis to get the significance. Likewise you could change the predictors vs. moderator status of both variables in PROCESS to get the conditional effects like you now have for your group IV.

HTH

Holger

Holger Steinmetz Yes, how SDO interacts with the relationship between experimental conditions and bystander intentions is my main focus here. SDO effects on the groups is more of a posthoc. But actually, I believe it is relevant as part of my theoretical basis for the research is social identity theory (specifically, does social identity theory influence participants behavioural intention toward an ingroup/outgroup). I ran an independent t-test prior, to see if there were group differences lay and it was not significant.