Error in multilevel path analysis (with moderated serial mediation)

[Laura]

Dear lavaan group,

I am amazed by what lavaan can do and really would like to get it to work for my model as well.

Right now, I am trying to do a multilevel path analysis, including a moderated serial mediation (using the index of moderated mediation proposed by Hayes, 2015).

This is my code:

multilevel$interaction <- (multilevel$moderator_c * multilevel$IV_c)

 

MLmod.med.serial <- '

level: 1

DV ~ (wc1)*IV_c + (wc2)*moderator_c + (wc3)*interaction + (wb1)*mediator1_c + (wb2)*mediator2_c

mediator1_c ~ (wa11)*IV_c + (wa12)* moderator_c + (wa31)*interaction

mediator2_c ~ (wa21)*IV_c + (wa22)* moderator_c + (wa32)*interaction + (wd)*mediator1_c

mediator1_c ~~ mediator2_c

 

level: 2

DV ~ (bc1)*IV_c + (bc2)*moderator_c + (bc3)*interaction + (bb1)*mediator1_c + (bb2)*mediator2_c

mediator1_c ~ (ba11)*IV_c + (ba12)*moderator_c + (ba31)*interaction

tlb_c ~ (ba21)*IV_c + (ba22)*moderator_c + (ba32)*interaction + (bd)*mediator1_c

mediator1_c ~~ mediator2_c

 

# defined parameters

wind.m1 := wa11 * wb1

bind.m1 := ba11 * bb1

wind.m2 := wa21 * wb2

bind.m2 := ba21 * bb2

wind.serial := wa11 * wd* wb2

bind.serial := ba11 * bd* bb2

windex1 := wa31 * wb1

windex2 := wa32 * wb2

bindex1 := ba31 * bb1

bindex2 := ba32 * bb2

windex.serial := wa31 * wd * wb2

bindex.serial := ba31 * bd * bb2

'

 

MLmod.med.serial.fit <- sem(MLmod.med.serial, multilevel, cluster = "team", estimator = "MLR", bootstrap = 1000)

 

And this is the error I keep getting:

Error in if (fx.delta < tol) { : missing value where TRUE/FALSE needed

I would be really grateful for your help!

Best regards,

Laura

[Yves Rosseel]

> MLmod.med.serial.fit <- sem(MLmod.med.serial, multilevel, cluster =
> "team", estimator = "MLR", bootstrap = 1000)

multilevel + bootstrap is not yet supported in lavaan (0.6-3).

If you remove the bootstrap= argument, and run it with verbose = TRUE,
what do you see?

Yves.

[Laura]

Thanks a lot for your quick answer!

If I do what you suggested, I get:

Fitting unrestricted (H1) model in group  1 
EM iter:   0  fx =  -8934.0914083798 
EM iter:   1  fx =  -4543.0880957699  fx.delta = 4391.00331261 
EM iter:   2  fx =  -3733.4423660440  fx.delta = 809.64572973 
EM iter:   3  fx =  -2969.1494916639  fx.delta = 764.29287438 
EM iter:   4  fx =  -2211.8537773779  fx.delta = 757.29571429 
EM iter:   5  fx =  -1456.8285984765  fx.delta = 755.02517890 
EM iter:   6  fx =   -702.7848069543  fx.delta = 754.04379152 
EM iter:   7  fx =     50.7617113116  fx.delta = 753.54651827 
EM iter:   8  fx =    804.0279350211  fx.delta = 753.26622371 
EM iter:   9  fx =   1557.1236039216  fx.delta = 753.09566890 
EM iter:  10  fx =   2310.1127167566  fx.delta = 752.98911283 
EM iter:  11  fx =   3063.0642529586  fx.delta = 752.95153620 
EM iter:  12  fx =   3816.3675000648  fx.delta = 753.30324711 
EM iter:  13  fx =   4572.9979835792  fx.delta = 756.63048351 
EM iter:  14  fx =   5377.2642435012  fx.delta = 804.26625992 
EM iter:  15  fx =                NA  fx.delta =        NA 

Error in if (fx.delta < tol) { : missing value where TRUE/FALSE needed

Best regards

Laura

[Yves Rosseel]

Thanks. Clearly, something goes wrong here, while estimating the
saturated model. Would you be able to send me (privately) the model and
the data (or just a snippet of the data, just enough to reproduce this)?

Thanks in advance,

Yves.

> --
> 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
> <mailto:lavaan+un...@googlegroups.com>.
> To post to this group, send email to lav...@googlegroups.com
> <mailto:lav...@googlegroups.com>.
> Visit this group at https://groups.google.com/group/lavaan.
> For more options, visit https://groups.google.com/d/optout.

[Laura]

Update: Issue solved!

 

For others who might encounter the same problem:

I tried to model a cross-level interaction and introduced a level 2-variable in the level 1-section. This doesn't work because the level 2-variable has no level 1-variance. For a cross-level interaction to work, one would need a random slope on level 1 instead which is not yet possible in lavaan. According to Yves, lavaan dev 0.6-4.1343 (or higher) will produce the following warning:

 

Warning messages:

1: In lav_data_full(data = data, group = group, cluster = cluster,  :

   lavaan WARNING:

     Level-1 variable "mod" has no variance at the within level.

     The variable appears to be a between-level variable. Please remove

     this variable from the level 1 section in the model syntax.

2: In lav_h1_implied_logl(lavdata = lavdata, lavsamplestats = lavsamplestats,  :

   lavaan WARNING:

     H1 estimation resulted in a within covariance matrix with (near)

     zero variances for some of the level-1 variables: mod

 

Thanks a lot for the great help and the quick response!

[Ramjanul Ahsan]

Hi Laura,

Can I get your final code? I am struggling with the coding to look into the Moderated mediation effect.

Thanks in advance.

Ramjan 

[Laura]

Hi Ramjan,

I would if it had worked. Back then it did not work for the model I was trying to build so I ended up using Mplus.

But quite some time has passed since then so maybe it's possible to do this in lavaan now?!

Fingers crossed for your analyses!

Laura

[Ramjanul Ahsan]

Hi Laura,

Thanks for getting back to me. No it didn't work. Thats why I was bit optimistic when I see your post. 

Let's see how it goes.

Thanks again.

Ramjan

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/d40d2d26-e874-452e-b6fd-bcf6b293dea2%40googlegroups.com.