3 replies to this topic

[LIUXIAOLIN]

Hi all,

 

I've gained really small stderr and CV% when using QRPEM algorithm, attached below.

Does it a true value? If not, is it possible to solve this problem?

 

Besides, I've tried FOCE-ELS with sandwich method, but the CV% of parameters cannot be generated. So I used  QRPEM algorithm. Is there any better idea to generate the CV%？

 

Looking forward for your suggestions.

 

 

 

BR,

Lamm.

Attached Thumbnails

• 

Edited by LIUXIAOLIN, 28 February 2024 - 06:53 AM.

[Simon Davis]

that is indeed very small - i've not seen that before but without seeing the rest of your data it's hard to say if it is true - it seems unlikely but we'd need more to investigate fully, if you're concerned about sharing data then you could email to support to have it looked at confidentially.

 

 SImon.

[LIUXIAOLIN]

that is indeed very small - i've not seen that before but without seeing the rest of your data it's hard to say if it is true - it seems unlikely but we'd need more to investigate fully, if you're concerned about sharing data then you could email to support to have it looked at confidentially.

 

 SImon.

 

Hi Simon,

 

Thanks for your quick response. 

 

I could not upload the project file in this website, and I don't know why. Anyway, I've sent the project file to you via email. Thank you!

 

 

BR,

Lamm.

[Simon Davis]

Hi Lamm, it looks like you've hit a known issue related to QRPEM engine [QC 15474] QRPEM engine underestimates SEs for particular models.  We have fixed this in the 8.5 release that will come out after Easter.

 

Until you upgrade to that release the advice is that, if after working  with QRPEM engine you get really tiny CV%;
- accept the final estimates
- then run bootstrap or FOCE ELS wtih 0 iterations (just to get SE)

 

  Simon.