Usermodel: some estimated ov variances are negative

[Sarah Kajee]

Hi all, 

Im newish to gSEM (and don't come from a genetics/stats background). I am running a number of different user models for one of my PhD chapters and I keep getting a number of warning messages in most of them (pasted below). For context. I am looking at two groups (brain volumes and a neurological disorder which has 10 different sub-phenotypes). When I ran a standard LDSC it showed that there is an inverse relationship between the brain volumes and sub-phenotypes of the disorder. i.e as the severity of the disorder increases, reduced brain volume is seen in specific regions. 

My question is if I should ignore these warning considering that there is an inverse relationship between the disorder and brain volumes?

Warning messages: 1: In usermodel(anthro_June1.1, estimation = "DWLS", model = CFAofEFA_June1.1, : A difference greater than .025 was observed pre- and post-smoothing in the genetic covariance matrix. This reflects a large difference and results should be interpreted with caution!! This can often result from including low powered traits, and you might consider removing those traits from the model. If you are going to run a multivariate GWAS we strongly recommend setting the smooth_check argument to true to check smoothing for each SNP. 2: In usermodel(anthro_June1.1, estimation = "DWLS", model = CFAofEFA_June1.1, : A difference greater than .025 was observed pre- and post-smoothing for Z-statistics in the genetic covariance matrix. This reflects a large difference and results should be interpreted with caution!! This can often result from including low powered traits, and you might consider removing those traits from the model. If you are going to run a multivariate GWAS we strongly recommend setting the smooth_check argument to true to check smoothing for each SNP. 3: In usermodel(anthro_June1.1, estimation = "DWLS", model = CFAofEFA_June1.1, : The unstandardized model produced the following warning: lavaan->lav_object_post_check(): some estimated ov variances are negative 4: In usermodel(anthro_June1.1, estimation = "DWLS", model = CFAofEFA_June1.1, : The standardized model produced the following warning: lavaan->lav_object_post_check(): some estimated ov variances are negative

[Sarah Kajee]

My model fit: 
      chisq df      p_chisq      AIC       CFI       SRMR

df 117.4064 42 4.657872e-09 165.4064 0.9851008 0.06495223