0

In order to linearize my model I use $\ln(x+1)$, so I come up with the following equation: $$ \ln(y+1) = b_0 + b_1 \ln(x_1 + 1) + b_2\ln(x_2+1) + \cdots + b_n\ln(x_n+1). $$ I hypothesize based on latent (unobserved variables), and as measures I use observed variables (or proxies) - the ones in the equation. My plan is to use SEM (Structural equation modeling) or/and CFA (confirmatory factor analysis) to validate my model.

In doing my SEM/CFA should I use the transformed variables $\ln(x+1)$ or as they are observed (just $x$)?

QuantIbex
  • 3,880
  • 1
  • 24
  • 42
Delyan Peyankov
  • 103
  • 9
  • Closely related: http://stats.stackexchange.com/questions/30728/how-small-a-quantity-should-be-added-to-x-to-avoid-taking-the-log-of-zero, http://stats.stackexchange.com/questions/41361/choosing-c-such-that-logx-c-would-remove-skew-from-the-population, and http://stats.stackexchange.com/questions/298/in-linear-regression-when-is-it-appropriate-to-use-the-log-of-an-independent-va. – whuber Jul 23 '13 at 16:26
  • How do you know if the model is linear in SEM before you've fitted it? It's generally OK to use the transformed variables in sem though. – Jeremy Miles Nov 27 '13 at 16:58

0 Answers0