1

Link to picture: https://www.dropbox.com/s/bxzxy0y37npqc4o/SEM%20problems.PNG

Need a bit of help determining what I should do to fix this problem: The SEM wont run because it believes it's non-positive definite. Meaning that there may be singularity.

However these items are very different as i have items measuring autonomy, job feedback, coworker support, positive mood and employee engagement.

Can anyone suggest a solution to this problem?

Gala
  • 8,323
  • 2
  • 28
  • 42
user28861
  • 11
  • 1
  • Whether you would expect the items to have different meaning is not terribly relevant, it can still happen especially with very small sample sizes. Previous questions on this: http://stats.stackexchange.com/questions/51473/matrix-is-not-positive-definite-even-when-highly-correlated-variables-are-re and http://stats.stackexchange.com/questions/8328/how-to-make-a-matrix-positive-definite http://stats.stackexchange.com/questions/52976/is-a-sample-covariance-matrix-always-symmetric-and-positive-definite – Gala Aug 06 '13 at 10:30

0 Answers0