Multicolinearity can be a cause; Heywood cases (negative variance) may happen; check the results; if this is the case, fix negative variance(s ) = a positive number close to 0 (say 0.005). Hope this helps.

Nguyen Tho Thank you for your comment Prof Tho. I’ve checked variances in Amos’s output & there is none negative one. The correlation between 2 IDVs is .65. Perhaps there is other reason?

Hello Linh,

Not to question your query, but, are you certain that you're looking at standardized coefficients? AMOS, by default, does not supply standardized coefficient estimates; they have to be specifically requested. See this advice from the IBM SPSS answer site (https://www-01.ibm.com/support/docview.wss?uid=swg21568925)

Standardized Estimates is [sic] not displayed by default within Amos. You will need to specify this output by opening the 'Analysis Properties' dialog window (Ctrl+A) and switching to the 'Output' tab.

Select the check box for 'Standardized estimates' and close the dialog window.

Since a new output has been selected, you will need to run the 'Calculate Estimates' (Ctrl+F9) on your Amos graphics file again in order for the 'Standardized estimates' to appear.

Good luck with your work!

Thank you Professor David Morse in fact I did exactly that.