[R] Negative variance with lavaan in a multigroup analysis.
Raphael Royaute
raphael.royaute at gmail.com
Mon Jun 4 18:21:23 CEST 2012
Hi list members,
I saw a couple lavaan posts here so I think I’m sending this to the
correct list.
I am trying to run a multigroup analysis with lavaan in order to
compare behavioural correlations across two populations. I’m following
the method suggested in the paper by Dingemanse et al. (2010) in
Behavioural Ecology.
In one of the groups, lavaan returns negative variance for one path
and I’m unable to obtain the corresponding R-square. As far as I know,
this is not due to measurement error so I’m wondering what else could
be causing this. Since the behavioral variables were standardized at
the population level prior to SEM analysis, lavaan should return 1 for
the variance and all paths coefficients should vary between 0 and 1,
right? Is there anything wrong in my code that may be causing this?
Thanks a lot for your help!
Raphael Royaute
PhD candidate
Dept Natural Resource Sciences
McGill University, Macdonald campus
Below is my code and output (my apologies, this is a pretty lengthy output):
Riskmodel2_cfa <- '# Latent variable
SySt =~ Activity+Aggression+Voracity
# Residuals
Boldness~~Boldness'
fit.riskmodel2_multi_Ad_cfa <- cfa(Riskmodel2_cfa, data=Ad_scale,
std.lv=T,std.ov=T,
se= "boot",test="bootstrap",
bootstrap=5000L,group="Site")
summary(fit.riskmodel2_multi_Ad_cfa,standardized=T, rsquare=T)
lavaan (0.4-14) converged normally after 50 iterations
Number of observations per group
1 56
0 41
Estimator ML
Minimum Function Chi-square 6.171
Degrees of freedom 6
P-value 0.404
P-value (Bollen-Stine Bootstrap) 0.566
Chi-square for each group:
1 4.832
0 1.339
Parameter estimates:
Information Observed
Standard Errors Bootstrap
Number of requested bootstrap draws 5000
Number of successful bootstrap draws 3423
Group 1 [1]:
Estimate Std.err Z-value P(>|z|) Std.lv Std.all
Latent variables:
SySt =~
Activity 0.295 0.193 1.532 0.126 0.295 0.298
Aggression 0.962 7.533 0.128 0.898 0.962 0.971
Voracity 0.481 1.223 0.393 0.694 0.481 0.485
Intercepts:
Activity 0.000 0.132 0.000 1.000 0.000 0.000
Aggression 0.000 0.131 0.000 1.000 0.000 0.000
Voracity -0.000 0.133 -0.000 1.000 -0.000 -0.000
Boldness 0.000 0.130 0.000 1.000 0.000 0.000
SySt 0.000 0.000 0.000
Variances:
Boldness 0.982 0.192 0.982 1.000
Activity 0.895 0.194 0.895 0.911
Aggression 0.056 291.974 0.056 0.057
Voracity 0.751 31.367 0.751 0.765
SySt 1.000 1.000 1.000
Group 2 [0]:
Estimate Std.err Z-value P(>|z|) Std.lv Std.all
Latent variables:
SySt =~
Activity 1.599 11.281 0.142 0.887 1.599 1.619
Aggression 0.162 0.219 0.740 0.459 0.162 0.164
Voracity 0.260 0.238 1.094 0.274 0.260 0.263
Intercepts:
Activity 0.000 0.155 0.000 1.000 0.000 0.000
Aggression -0.000 0.154 -0.000 1.000 -0.000 -0.000
Voracity 0.000 0.156 0.000 1.000 0.000 0.000
Boldness 0.000 0.156 0.000 1.000 0.000 0.000
SySt 0.000 0.000 0.000
Variances:
Boldness 0.976 0.156 0.976 1.000
Activity -1.580 459.764 -1.580 -1.620
Aggression 0.949 0.197 0.949 0.973
Voracity 0.908 0.248 0.908 0.931
SySt 1.000 1.000 1.000
R-Square Group 1:
Activity 0.089
Aggression 0.943
Voracity 0.235
R-Square Group 0:
Activity NA
Aggression 0.027
Voracity 0.069
More information about the R-help
mailing list