jamovi

Hello, I am not sure if I am going mad or not, but I think that a column has disappeared from the SEM results tables for model fit indices.

I am attempting to do a CFA via the SEM module so I could have access to a different range of fit indices. I worked from the same file for a couple of weeks, duplicating analyses and slightly modifying the model each time. I came back to this a couple of days ago after a break and I feel like something is missing. I had been looking at the tables shown in the attached image, and I could swear that last week they also had a column headed 'scaled', which was the one I was using, and I had noted this in my own notes (which is why I think I might not be completely mad....). Often, the results shown for CFI and TLI were 1 in the 'model' column (as you can see here), but I was ignoring that in favour of what looked like more realistic results from a second column headed 'scaled', which were around the 0.95+ mark each time. Now that column is gone and all I have is the suspiciously good results shown here in a single column.

My data is ordinal from Likert scales, so I am using the Robust Weighted Least Squares (WLSMV) method. This 'lost column' has happened the table titled 'User model vs baseline model' and the one below that called 'Additional Fit Indices' for all my different analyses. The other tables above that, titled 'model tests' and 'Fit Indices' (showing Chi sq, RMSEA and SRMR) still contain scaled results as well as basic user model results.

Any ideas what has happened?

just press F5 to update the results

It's not that it's not updating the results, it's that the rendering of the tables has changed such that an entire column is just no longer there. So there's no column heading that says 'Scaled'. I feel sure there was a column to the right of the 'Model' column which said 'Scaled' and gave slightly different results, not least because in my notes I have referred to that, and none of the results I have noted down were a perfect '1' for CFI or TLI, which is what shows in the basic 'Model' column in the image below.

So pressing F5 refreshes the system amd updates all the other tables and results, as expected, but the column header itself for this data is gone, and there is nothing suggesting it was ever there or ever calculating anything.

I really do feel sure it was there. I know that the SEM module is based on Lavaan but unfortunately I don't have the skills to use r directly to see if this results column would be given to me another way.

Just an update, I have been looking at the lavaan.org pages to see if I can find anything there and there is a discussion from the 0.5-22 release that suggests that scaled and/or robust estimates of the CFI, TLI etc are/were available in lavaan itself, which makes me think I am not imagining that they would have also been printed out in the tables on the Jamovi SEM module if it comes from Lavaan originally?

"Version 0.5-22
Released on CRAN: 24 September 2016
New features and user-visible changes:
(old) ‘scaled’ versions of CFI/TLI/RMSEA are again printed in the summary() output
the (new) robust versions of CFI/TLI/RMSEA are printed on separate lines"

From https://lavaan.ugent.be/history/dot5.html

Ok, the issue is clear. We sort it out in next version. Thx for reporting it with details.

Thank you. Do you have any idea of a timeline for that update? I need to hand something in to my supervisor in about 10 days and will have to explain the delay to him and he’ll probably ask when I can pick it up again. Would be really useful to have an indication I can give him. Many thanks for your help on this.

I cannot reproduce the issue. If I run a model with "WLSMV" estimation method, I get all the required columns in the table you mentioned (scaled and robust). Could you share a omv file with the issue, so we can check what may be going on?

Hi, I will send an omv file to you by email. FYI this issue is happening on the desktop version, which I use because the cloud version seems to make my laptop very tired and the fan runs all the time. The desktop version I am using is 2.3.28.

Yesterday I used the cloud version to see what would happen, and opened the file, and as it was updating these missing columns just re-appeared. But when I open the file in the desktop version they are still not there. So I can at least get on with some work by using the cloud version, but it still seeems odd still that this happens in the desktop environment.

In the omv file I also deleted all the analyses except the first, one as I wondered if it were to do with me having duplicated the thing so many times (11 in total I think), making it too big a file or something. But that hasn't changed anything, even with one analysis in there, the columns are still missing.

Thanks for looking into this more.

If you update jamovi and semlj to the latest version, you ll find all the columns yiu are missing

Hi, I have installed the most up to date destop version of Jamovi and the Semlj but now all of my analyses are greyed out and they have a warning at the top which says 'condition has length <1'.

They still work on the cloud version. I will just keep trying with that.