Can a mediator change the direction of the relationship between variables?

[Daniela L]

Hello All, 

Age is directly associated with the tendency to cease driving; when people grow older, the tendency to stop driving increases (a positive path). Also, age influences cognitive, visual, and health-related tests. It takes longer for you to complete a task test as you get older (a positive path).  And the longer task test time related to a higher tendency of driving cessation (positive path). 

After I added the cognitive and visual task completion time in the model as a meditator of age on driving cessation, the direction of the direct path between age and giving up driving changed to negative. Now the results are not interpretable (The older a person is, the less likely they are to stop driving, no...). 

                                        Ets        SE        Z-value

Visual         ~      Age   0.278  0.043  6.474

Cognitive   ~       Age    0.432   0.057  7. 631

Give up driving ~ Age   -0.975  0.182   -5.353     

                              Visual  1.589  0.082   19.328

                              Cognitive 2.133  0.120  17.785

                               Health  2.465    0.076      32.282

Can a moderator or mediator affect the direction of the relationship between variables? how to explain such results.

Many thanks.

Dan

[Daniela L]

Sorry, please ignore the attached pdf files. 

Daniela L <liangdan...@gmail.com> 于2021年6月17日周四 下午10:20写道：

--
You received this message because you are subscribed to the Google Groups "lavaan" group.
To unsubscribe from this group and stop receiving emails from it, send an email to lavaan+un...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/lavaan/e88eadc9-f06f-4e51-827c-3ca813c022c5n%40googlegroups.com.

[Daniela L]

Please ignore the attached files, sorry for bothering.

[car...@web.de]

I assume that you mean the sign. That is conceivable. But I don't see a mediator in your model. Only direct effects of age on 3 DV?

--
Diese Nachricht wurde von meinem Android Mobiltelefon mit WEB.DE Mail gesendet.

Am 18.06.21, 04:24 schrieb Daniela L <liangdan...@gmail.com>:

--

You received this message because you are subscribed to the Google Groups "lavaan" group.
To unsubscribe from this group and stop receiving emails from it, send an email to lavaan+un...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/lavaan/a18b4557-1c7f-491c-a634-35e4091dbdeen%40googlegroups.com.

[Pat Malone]

C, 

I believe Visual and Cognitive mediate the effect of Age on Give Up Driving.

But yes, it's entirely possible. We rarely know the full set of in-the-state-of-the-world mediators between X and Y. The "direct effect" (I prefer the term "residual effect," though "residual" already has plenty of meanings), is simply the action of X on Y via the universe of mediators of the relation minus the two included in the model. Some of them may be inconsistent mediation, which is certainly a thing. (Though if the direct/residual effect is not statistically significant, I wouldn't worry about trying to interpret it).

A side note--if that's the only place Health shows up in the model, I would suspect some specification error--e.g., the absence of Cognitive ~ Health.

Hope this helps.

To view this discussion on the web visit https://groups.google.com/d/msgid/lavaan/trinity-2c1ace95-5244-4228-9da9-fd3015d4f7fd-1624004033948%40msvc-mesg-web003.

--

Patrick S. Malone, PhD

Sr Research Statistician, FAR HARBΦR

+1 803.553.4181  |  pat@  |  farharbor.com

This message may contain confidential information; if you are not the intended recipient please notify the sender and delete the message.