covariance matrix, missing variables

[ju.lis.br...@gmail.com]

Hi,

I fitted two models using sem(). Both models run  perfectly fine and summary() returns estimates for all variables. Using lavInspect(model1, what="cov.ov") I get a covariance matrix from model 1 in which the exogenous categorical variables landscapeforest.fragment, landscapeisland, sexM and season2014.2015 (all are dummy coded) are missing. Other categorical variables (supertypes3x1, etc.) do however appear in the covariance matrix. In model 2 the exogenous variables landscapeforest.fragment and landscapeisland are present in the covariance matrix. 

Is there something wrong in my script or is there another reason for this? See below for the script:

Thank you very much for your help!

Julian

data2[,c("tryp.inf0",

        "n1.2",

        "n15.29",

        "n23",

        "hepaci.inf",

        "picorna",

        "picobirna",

        "supertype3x1",

        "supertype5x1",

        "supertype6x1",

        "supertype9x1",

        "supertype10x1")] <-

  lapply(data2[,c("tryp.inf0",

                 "n1.2",

                 "n15.29",

                 "n23",

                 "hepaci.inf",

                 "picorna",

                 "picobirna",

                 "supertype3x1",

                 "supertype5x1",

                 "supertype6x1",

                 "supertype9x1",

                 "supertype10x1")], ordered)

model1  = "#regressions

#direct effects

hepaci.inf ~ landscapeforest.fragment + landscapeisland + sexM + supertype1x1

picobirna ~ season2014.2015

picorna ~ a*juliDensity + s*season2014.2015 + t*supertype10x1 + u*supertype3x1 + v*supertype5x1 + w*supertype9x1

tryp.inf0 ~ b*marsu.density + landscapeforest.fragment + mean_pdist

n1.2 ~ d*SimpsonInv2018ed.in + c*juliDensity + landscapeisland

n15.29 ~ f*SimpsonInv2018ed.in + landscapeisland + e*juliDensity + x*supertype6x1

n23 ~ h*SimpsonInv2018ed.in + g*juliDensity 

SimpsonInv2018ed.in ~ k*landscapeforest.fragment

juliDensity ~ i*landscapeisland

marsu.density ~ j*landscapeforest.fragment + season2014.2015

supertype3x1 ~ l*season2014.2015 + m*landscapeisland

supertype5x1 ~ n*landscapeisland

supertype6x1 ~ o*landscapeforest.fragment + p*landscapeisland

supertype9x1 ~ q*landscapeisland

supertype10x1 ~ r*season2014.2015

#indirect effects 

#landscapeisland~picorna via juliDensity

ai := a*i

#landscapeforest.fragment~tryp.inf0 via marsu.density

bj := b*j

#landscapeforest.fragment/landscapeisland~n1.2 via juliDensity,SimpsonInv2018ed.in

ci := c*i

dk := d*k

#landscapeforest.fragment/island~n15.29 via juliDensity, SimpsonInv2018ed.in

ei := e*i

fk := f*k  

#landscapeisland/forest.fragment~n23 via juliDensity, SimpsonInv2018ed.in

gi := g*i

hk := h*k

#season2014.2015/landscapeisland~picorna via st3

sl := s*l

um := u*m

#landscapeisland~picorna via st5

vn := v*n

#landscapeforest.fragment/landscapeisland~n15.29 via st6

xo := x*o

xp := x*p

#landscapeisland~picorna via st9

wq := w*q

#season2014.2015~picorna via st10

tr := t*r"

model1 = sem(model1, data2, estimator = "WLSMV")

lavInspect(model1, what = "cov.ov")

	hepaci.inf	picobirna	picorna	tryp.inf0	n1.2	n15.29	n23	SimpsonInv	juliDensity	marsu.density	supertype3x1	supertype5x1	supertype6x1	supertype9x1	supertype10x1
hepaci.inf	1.000
picobirna	-0.305	1.000
picorna	0.000	0.185	1.367
tryp.inf0	-0.077	0.319	0.052	1.000
n122	0.071	-0.089	-0.200	-0.134	1.000
n15.29	0.065	0.082	0.287	0.166	0.196	1.066
n23	0.045	0.155	0.176	-0.030	0.162	-0.088	1.000
SimpsonInv	0.000	0.000	0.000	0.000	-0.198	-0.129	-0.122	0.176
juliDensity	0.000	0.000	1.148	0.000	0.718	1.358	-0.971	0.000	17.061
marsu.density	0.000	0.000	0.000	0.125	0.000	0.000	0.000	0.000	0.000	0.213
supertype3x1	0.000	0.000	0.192	0.000	0.000	0.000	0.000	0.000	0.000	0.000	1.000
supertype5x1	0.000	0.000	0.185	0.000	0.000	0.000	0.000	0.000	0.000	0.000	0.000	1.000
supertype6x1	0.000	0.000	0.000	0.000	0.000	0.258	0.000	0.000	0.000	0.000	0.000	0.000	1.000
supertype9x1	0.000	0.000	-0.329	0.000	0.000	0.000	0.000	0.000	0.000	0.000	0.000	0.000	0.000	1.000
supertype10x1	0.000	0.000	0.433	0.000	0.000	0.000	0.000	0.000	0.000	0.000	0.000	0.000	0.000	0.000	1.000

model 2 = "#regressions

#direct effects

multi.inf ~ landscapeforest.fragment + b*SimpsonInv2018ed.in + a*juliDensity 

SimpsonInv2018ed.in ~ d*landscapeforest.fragment

juliDensity ~ c*landscapeisland

#indirect effects

#landscapeisland~multi.inf via juliDensity

ac := a*c

#landscapeforest.fragment~multi.inf via SimpsonInv2018ed.in

bd := b*d"

 

sem(model2, data2, estimator = "MLM")

lavInspect(model2, what = "cov.ov")

	multi.inf	SimpsonInv	juliDensity	landscapeforest.fragment	landscapeisland
multi.inf	1.149
SimpsonInv	-0.218	0.399
juliDensity	1.587	-0.551	24.250
landscapeforest.fragment	-0.178	0.196	-0.536	0.191
landscapeisland	0.142	-0.105	1.254	-0.103	0.240

[Terrence Jorgensen]

In model 2 the exogenous variables landscapeforest.fragment and landscapeisland are present in the covariance matrix. 

Is there something wrong in my script or is there another reason for this?

You are using 2 different estimators, which may have different default settings for the fixed.x argument (see the ?lavOptions help page for a description, although it will not say what controls the default setting).  You can see what the options are:

lavInspect(model1, what = "options")$fixed.x
lavInspect(model2, what = "options")$fixed.x

FYI, if they are exogenous, do NOT define the dummy codes as ordered.  They should be numeric dummy codes, and set fixed.x=TRUE to avoid making distributional assumptions about them (i.e., they don't have to be multivariate normal because their summary statistics are used instead of actually estimating their (co)variances).

http://lavaan.ugent.be/tutorial/cat.html

Terrence D. Jorgensen

Assistant Professor, Methods and Statistics

Research Institute for Child Development and Education, the University of Amsterdam

http://www.uva.nl/profile/t.d.jorgensen