lavaan WARNING: covariance matrix of latent variables but lavInspect( fit, "cov.lv") are all positive,what's the problem?

[ss xx]

> df_com_cfa<-lavaan::cfa(lavaan_model,df_com,ordered=colnames(df_com))
Warning message:
In lav_object_post_check(object) :
  lavaan WARNING: covariance matrix of latent variables
                is not positive definite;
                use lavInspect(fit, "cov.lv") to investigate.
> lavInspect( df_com_cfa, "cov.lv")
    f1    f2    f3    f4    f5    f6    f7    f8    f9    f10  
f1  0.247                                                      
f2  0.311 0.502                                                
f3  0.329 0.494 0.679                                          
f4  0.379 0.590 0.595 0.727                                    
f5  0.377 0.531 0.616 0.682 0.725                              
f6  0.359 0.494 0.567 0.620 0.659 0.629                        
f7  0.326 0.453 0.524 0.568 0.597 0.561 0.704                  
f8  0.301 0.461 0.470 0.554 0.547 0.494 0.509 0.513            
f9  0.327 0.475 0.537 0.575 0.588 0.538 0.538 0.490 0.570      
f10 0.275 0.441 0.459 0.525 0.472 0.437 0.465 0.444 0.462 0.481

Covariance matrix are all positive,how to solve this  problem?

[car...@web.de]

?

Am 29.06.20, 21:15 schrieb ss xx <mailtoy...@gmail.com>:

--
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 view this discussion on the web visit https://groups.google.com/d/msgid/lavaan/608fab3f-c9c8-4ced-b746-ccb9da3e48dco%40googlegroups.com.

[Patrick (Malone Quantitative)]

"Not positive definite" does not refer to the signs of the
covariances. From looking at this, I'd guess the culprit is a
standardized covariance (i.e., correlation) >1.

> --
> 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 view this discussion on the web visit https://groups.google.com/d/msgid/lavaan/608fab3f-c9c8-4ced-b746-ccb9da3e48dco%40googlegroups.com.

--
Patrick S. Malone, Ph.D., Malone Quantitative
NEW Service Models: http://malonequantitative.com

He/Him/His

[ss xx]

Thanks for your reply.  Any problem with correlation?

> lavInspect( df_com_cfa2, "cor.lv")
    f1    f2    f3    f4    f5    f6    f7    f8    f9    f10  
f1  1.000                                                      
f2  0.885 1.000                                                
f3  0.803 0.847 1.000                                          
f4  0.896 0.977 0.847 1.000                                    
f5  0.891 0.881 0.877 0.939 1.000                              
f6  0.911 0.879 0.868 0.918 0.975 1.000                        
f7  0.782 0.761 0.758 0.794 0.835 0.843 1.000                  
f8  0.845 0.909 0.796 0.908 0.898 0.870 0.846 1.000            
f9  0.873 0.888 0.862 0.893 0.915 0.899 0.849 0.907 1.000      
f10 0.797 0.898 0.804 0.888 0.799 0.795 0.799 0.894 0.882 1.000

Patrick (Malone Quantitative)於 2020年6月30日星期二 UTC+8上午3時45分16秒寫道：

[gbr...@aucklanduni.ac.nz]

I agree with Patrick. r=.977 between f2 and f4. it's not over 1.00 but being so close to 1.00 suggests that those two factors measure ONE thing not 2. so for me i'd merge those 2 factors and run again. Of course you will need to explain why f2 and f4 are so closely aligned.

[Patrick (Malone Quantitative)]

Adding that your factor correlations are very high across the board.
The lowest one I spot is .76.

You might need to return to the drawing board in terms of how your CFA
reflects your theory and measurement.

Pat

> --
> 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 view this discussion on the web visit https://groups.google.com/d/msgid/lavaan/e3db4bf4-0d6b-4aff-8ae8-d2272c6676f9n%40googlegroups.com.