Difference between sem and multi regression analyszis
Answers
Answer:
Multiple regression is observed-variable (does not admit variable error), whereas SEM is latent-variable (models error explicitly).
Step-by-step explanation:
on a fundamental level and from a historical perspective, regression is most and above all a data-focused technique to place a line/plane in a multidimensional scatter plot. The fact that it is used by researchers to test causal hypotheses does not change the effect - even if you use regression estimates as a representation of your assumed effect.
SEM in contrast is a reflection of your underlying causal beliefs which consist in "weak assumptions" (the effects) and "strong assumptions" (belief about non-effects). Both create a causal structure which has implications for a certain data/correlation pattern. This implication a) creates a chance to test the structure (by means of model-data fit) and b) is involved in the estimation of the single parameters.
Of course, a SEM can be more or less saturated vs. restricted (sparse). As an extreme, you may have one dependent variable but several exposure variables. In this case, the SEM becomes statistically identical to the regression.