I ran CFA in R on ordinal data using DWLS and encountered a covariance error. My revised code is:
mod_new<-'F1=~a1+a2;F2=~a3+a4'
How can I fix this?
I encountered something similar when working with ordinal data in CFA. I found that it was essential to ensure that my data was appropriately coded as ordinal and that the estimation method was correctly specified in the model call. After verifying the model’s structure, I pre-processed the data to confirm that the ordinal levels were accurately interpreted. Experimenting with alternative estimation options and fine-tuning the starting parameter values helped resolve the covariance error in my case. A detailed review of both the data and model specifications can often highlight the underlying issue.
hey, try checkin if ur factors are indetified correctly. sometimes switching to a polychoric corr helps fix the cov err. also check ur starting vals and tolerances settings. best luck!
Hey there! I’ve been tinkering with similar models and found that sometimes the covariance error you’re seeing can be tied back to how the thresholds for ordinal items are estimated. Even if the data are in the right format, the solver sometimes gets thrown off by unexpected threshold behavior. I ended up trying a simplified model first to see if one of the items was causing the issue, then gradually added complexity back in. Have you tried isolating parts of the model or tweaking the threshold settings? I’m really curious, because it might lead to a more pinpointed source of the error. Let me know what you discover or if you’ve tried any other methods to get around this error! 
hey, try rechecking your item loadings and tweaking threshold settings. in my experimnt, reordering items and slight parameter adjustments helped resolve a similar cov err. sometimes minor tweaks make all the diff.
I faced a similar issue when running CFA under non-normal conditions with ordinal data. After careful investigation, I realized that ensuring proper treatment of thresholds was crucial. Declaring the variables as ordinal explicitly and verifying that the right type of correlation, such as polychoric correlations, was implemented helped resolve the covariance error. I also adjusted convergence criteria and checked for any irregularities in item coding. These steps, applied progressively, helped me isolate and fix the problem. I recommend reviewing both the measurement model and the handling of ordinal data.