Can Large Language Models Build Causal Comparativequasi Experimental