I am conducting what amounts to several simultaneous regressions on a latent variable. I'm examining whether hypothesized protective factors (hopefulness, parenting capacity, and educational environment) moderate the relationship between trauma exposure and externalizing symptoms in children. I can copy my syntax if needed, but it is too long to include in this message.
My question is this: I originally ran the analyses with my three hypothesized protective factors coded normally (on a scale from 0-3, with 3 indicating poorer scores on a particular factor). This gave me some significant results, but the correlations are counter-intuitive. I then recoded the variables using the syntax below, but when it ran it made all of my path coefficients non-significant. To my knowledge, this should not happen - it should just change the sign of the coefficients. Anyone have any insight as to why this might be happening? Is it something to do with my syntax or data, or is it a problem with my copy of Mplus?