my question is a rather theoretical one. I have to decide in advance how I want to analyze my data (I’m going with the lme4 package in R) and feel torn between doing a model comparison by creating two nested models (m1) that differ in one fixed effect (m2) and just running the full mixed model (m2). Either way, I would look mainly at the p-value (either of the output of anova(m1, m2) or of the beta coefficient of the fixed effect in m2 that I am most interested in). Does anyone know if it is even possible to obtain different p-values for the two different approaches? If the results are always the same anyway (which is my experience so far), what makes you decide for or against one or the other approach? I have seen both in psychological papers and feel confused about it. I would really appreciate it if anyone can share their experience or line of thought regarding those two approaches.
Thank you so much for reading this post so far.

I have six conditions. We can call them A, B, C, D, E, F.
I am trying to create a contrast matrix in R. I will use these contrasts in a regression model.