To use CFA, you need to understand the research question that you are addressing. Are you validating scales (or subscales) based on a theoretical model or framework? Has there been previous work using EFA that established the scales or constructs you're examining? If your answer is "yes" to both questions, go ahead and use CFA. As for WLS, which I presume to be the "arbitrary generalized linear squares" (AGLS) method, you apply it when there isn't normality in your dataset (which happens regularly), The good thing about WLS/AGLS is that you can apply it to continuous, binary, and ordinal variables. However, the pitfall is that you need a considerably large sample size. The sample needs to be larger than the number of elements in your weight matrix, which pertains to [Sq(p*p+1)] / 2. For example, if you have 6 variables, you would need a preferred sample size of 113.
thank you for the answer. If i would adjust my size to 113, wouldnt it affect the result of CFA since i also took into consideration the proper sample size with regrds to CFA