Using ML and DWLS estimators in ordinal data (efa in lavaan)

[Henrique Oliveira]

HI,

I have a set of 84 answers to a questionnaire, with a total of 14 questions. The answers to all the questions use a 4-point Likert scale (1 to 4).
As the answers are on an ordinal scale, I'm using the DWLS estimator (as recommended) to perform an EFA analysis (1 factor only), using lavaan's 'efa' function. As a result, I'm getting the following results: p-value = 1, CFI = 1 and RMSEA = 0.
However, if I use the 'ML' estimator, the result I get is: p-value = 0, CFI = 0.855 and RMSEA = 0.148.
Can anyone help me understand the reason for the strange values of p-value, and CFI with the DWLS estimator, which is more suitable for ordinal scale data than the ML estimator?

Many thanks

Henrique