How to do meta-analysis with two moderators simultaneously?

[凌楚定]

Dear list,

Now I have a question about the meta-analysis with two moderators simultaneously. For sake of illustration, I use de Wit et al. (2012) paper as an example, which is similar to what I want to test. In this paper, the authors tested multiple moderators’ effects on task conflict-performance relationship with the weighted least squares multiple regression analysis (See Table 6 in the paper).

Slightly departing from what they did, what I want to test is a WLS regression model with only two moderators. One is a categorical variables, say, research setting; the other is a continuous variables, say, the association between task conflict and relationship conflict. Take the paper above as an example, what I want to explore is the effect size of the task conflict-performance relationship under the 2 (association: high vs. low) × 2 (research setting: field vs. laboratory study) conditions. Besides, I also want to plot the moderating pattern like Figure 2 in the paper.

A plausible solution is as follows. In addition to entering the two moderators simultaneously when running the WLS regression, I also include the product term of the two moderators. In other words, I formulate the WLS regression model just as the way when we test a general moderation/interaction effect model. After this, I plot the moderation pattern and do simple slope test. If the simple test shows both of the two lines are significantly different from zero, then we can make a conclusion like this: In the field study research setting, the association between task conflict and group performance varies as a function of the association between task conflict and relationship conflict, and when the association between task conflict and relationship conflict is higher, the relationship between task conflict and group performance is stronger (just for example, I have not run the data and I am not sure what I will get)…

I am not sure the solution above is correct or not. I will be grateful if any of you could give me some suggestions on this procedure or other alternative solutions. Thanks!

The paper I take as an example is: de Wit, F. R., Greer, L. L., & Jehn, K. A. (2012). The paradox of intragroup conflict: a meta-analysis. Journal of Applied Psychology, 97(2), 360.

---

​Chu-Ding Ling
Ph.D. Student of Business Administration
School of Management
Zhejiang University

===================== To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD

[凌楚定]

Thanks for Rich's reply in comp.soft-sys.stat.spss. I get some helpful information from another group. The following is the information the guy provided to me: http://meta-analysis.ning.com/forum/topics/interaction-between-two-moderators?xg_source=activity

If you find some other alternative solutions, please feel free to share with us.

Chu-Ding Ling

2014-07-13 20:41 GMT+08:00 Chu-Ding Ling <[hidden email]>:

Dear list,

Now I have a question about the meta-analysis with two moderators simultaneously. For sake of illustration, I use de Wit et al. (2012) paper as an example, which is similar to what I want to test. In this paper, the authors tested multiple moderators’ effects on task conflict-performance relationship with the weighted least squares multiple regression analysis (See Table 6 in the paper).

Slightly departing from what they did, what I want to test is a WLS regression model with only two moderators. One is a categorical variables, say, research setting; the other is a continuous variables, say, the association between task conflict and relationship conflict. Take the paper above as an example, what I want to explore is the effect size of the task conflict-performance relationship under the 2 (association: high vs. low) × 2 (research setting: field vs. laboratory study) conditions. Besides, I also want to plot the moderating pattern like Figure 2 in the paper.

A plausible solution is as follows. In addition to entering the two moderators simultaneously when running the WLS regression, I also include the product term of the two moderators. In other words, I formulate the WLS regression model just as the way when we test a general moderation/interaction effect model. After this, I plot the moderation pattern and do simple slope test. If the simple test shows both of the two lines are significantly different from zero, then we can make a conclusion like this: In the field study research setting, the association between task conflict and group performance varies as a function of the association between task conflict and relationship conflict, and when the association between task conflict and relationship conflict is higher, the relationship between task conflict and group performance is stronger (just for example, I have not run the data and I am not sure what I will get)…

I am not sure the solution above is correct or not. I will be grateful if any of you could give me some suggestions on this procedure or other alternative solutions. Thanks!

The paper I take as an example is: de Wit, F. R., Greer, L. L., & Jehn, K. A. (2012). The paradox of intragroup conflict: a meta-analysis. Journal of Applied Psychology, 97(2), 360.

---

​Chu-Ding Ling
Ph.D. Student of Business Administration
School of Management
Zhejiang University

===================== To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD