correlation between exogenous variable

[oumaima mtaallah]

Dear all,

I am new to SEM and lavaan.

In my model, I have four exogenous variables and I would like to add correlation between some of them.

The script is below :

model5='
#latent variable
Health_L =~ health+ physical_health+ time_hosp+ night+ annuel_visi_cate
# Regression operator
Health_L ~ a*gender+n*marital_status+e*age+m*owners+i*nationality+k* income
#Meditor
marital_status ~ d*age+ g*nationality + b*gender
owners ~ c*gender+f*age+ h*nationality+ l*marital_status+ j*income
# covariance
health ~~ physical_health
physical_health ~~ time_hosp
physical_health ~~ night
health ~~ time_hosp
health ~~ night
time_hosp ~~ night

#delete correlation
income ~~ 0*nationality # No covariance between income and nationality
gender ~~ 0*nationality #No covariance between gender and nationality
age ~~ 0*gender  #No covariance between gender and age
age ~~ 0*nationality

#Correlation

income~~ nationality
income~~ age
income~~ gender
#indirect effect:
# Marital_status:
bn:=b*n # Gender
dn:=d*n #Age
gn:=g*n #Nationality

#Owners:
cm:= c*m #gendre
fm:= f*m #Age
hm:= h*m #nationality
jm:= j*m #Income

#matial_status/owners:
blm:= b*l*m #gender
dlm:= d*l*m #Age
glm := g*l*m #nationality

#Total effect
total:= a+e+n+m+i+k+(b*n)+(d*n)+(g*n)+(c*m)+(f*m)+(h*m)+(j*m)+(b*l*m)+(d*l*m)+(g*l*m)
'

Warning messages:
1: In lav_data_full(data = data, group = group, cluster = cluster,  :
  lavaan WARNING: some observed variances are (at least) a factor 1000 times larger than others; use varTable(fit) to investigate
2: In muthen1984(Data = X[[g]], ov.names = ov.names[[g]], ov.types = ov.types,  :
  lavaan WARNING: trouble constructing W matrix; used generalized inverse for A11 submatrix

varTable(data_comp)
                 name idx  nobs    type exo user        mean          var nlev      lnam
1                   X   1 22351 numeric   0    0  272777.493 1.610155e+10    0          
2                 hid   2 22351 numeric   0    0  254213.133 3.321329e+10    0          
3                 pid   3 22351 numeric   0    0 2486966.875 3.435039e+12    0          
4                 pnr   4 22351 numeric   0    0       1.760 1.051000e+00    0          
5          year_birth   5 22351 numeric   0    0    1954.827 2.993560e+02    0          
6              gender   6 22351 numeric   0    0       1.520 2.500000e-01    0          
7         nationality   7 22351 numeric   0    0       1.092 8.400000e-02    0          
8      marital_status   8 22350 numeric   0    0       0.617 2.360000e-01    0          
9               horse   9 22351 numeric   0    0       0.011 1.100000e-02    0          
10                dog  10 22351 numeric   0    0       0.148 1.260000e-01    0          
11                cat  11 22351 numeric   0    0       0.171 1.420000e-01    0          
12              birds  12 22351 numeric   0    0       0.077 7.100000e-02    0          
13               fish  13 22351 numeric   0    0       0.063 5.900000e-02    0          
14             others  14 22351 numeric   0    0       0.090 8.200000e-02    0          
15               none  15 22186 numeric   0    0      13.158 1.163490e+02    0          
16 employement_status  16 22351 ordered   0    0          NA           NA    5 1|2|3|4|5
17             income  17 21971 numeric   0    0       7.491 5.031000e+00    0          
18        trainig_req  18 21986 numeric   0    0       0.206 4.270000e-01    0          
19             health  19 22320 ordered   0    0          NA           NA    5 1|2|3|4|5
20    physical_health  20 22277 ordered   0    0          NA           NA    3     1|2|3
21   legally_handicap  21 22256  factor   0    0          NA           NA    2       1|2
22    handicap_extend  22 22289 numeric   0    0       6.373 4.029230e+02    0          
23       visit_doctor  23 22351 numeric   0    0       2.677 1.975800e+01    0          
24         visit_hosp  24 22282 numeric   0    0       1.879 1.060000e-01    0          
25        N_time_hosp  25 22313 numeric   0    0       0.162 2.950000e-01    0          
26       N_night_hosp  26 22342 numeric   0    0       1.752 6.967400e+01    0          
27          summation  27 22351 numeric   0    0      13.622 1.027970e+02    0          
28             owners  28 22186 numeric   0    0       0.402 2.400000e-01    0          
29                age  29 22351 numeric   0    0      46.173 2.993560e+02    0          
30           handicap  30 22289  factor   0    0          NA           NA    3     0|1|2
31          time_hosp  31 22309 ordered   0    0          NA           NA    3     0|1|2
32              night  32 22341 ordered   0    0          NA           NA    3     0|1|2
33       annuel_visit  33 22351 numeric   0    0      10.708 3.161330e+02    0          
34   annuel_visi_cate  34 22351 numeric   0    0       1.037 6.640000e-01    0          
35           age_cate  35 22351 ordered   0    0          NA           NA    5 0|1|2|3|4

I don´t know how to fix it ?

Any suggestions are welcome

Thank you so much.

Best regards

[Terrence Jorgensen]

I don´t know how to fix it ? 

You can multiply/divide by factors of 10 (or 100, or 1000) to make the variables' scales more similar. 

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

[oumaima mtaallah]

Thanks Terrence for you reply.

I used the function " scale" and I don´t have anymore warning message.

Do you think that is the same as if I divide the factor as you recommanded ?

Thank you a lot !

--
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 post to this group, send email to lav...@googlegroups.com.
Visit this group at https://groups.google.com/group/lavaan.
To view this discussion on the web visit https://groups.google.com/d/msgid/lavaan/955ba816-8d8c-4a74-ab49-c775423cf062%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Terrence Jorgensen]

I used the function " scale" and I don´t have anymore warning message. 

Do you think that is the same as if I divide the factor as you recommanded ? 

Definitely not.  scale() divides by a sample SD, so you end up analyzing correlations instead of covariances, which leads to incorrect SEs and test statistics.

https://pdfs.semanticscholar.org/3daa/5151f6a0931e9b6587cf06571486013f56ec.pdf

A sample statistic is not a standard because it will vary from sample to sample, so dividing by a sample SD should not be called "standardizing" (although almost all textbooks in social sciences do so).  A more accurate term would be "studentizing".  Dividing by a factor of 10^n is a standard that would apply to any sample, because your interpretation of slopes would simply change to "per 10^n units", and your SEs would change by the same factor, so your tests would be unaffected.

[oumaima mtaallah]

Thank you for the feedback, Terrence

Oumaima

--
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 post to this group, send email to lav...@googlegroups.com.
Visit this group at https://groups.google.com/group/lavaan.

To view this discussion on the web visit https://groups.google.com/d/msgid/lavaan/50ae5781-d337-4e59-a7d6-3b6bdafbcb31%40googlegroups.com.