Orthagonal Contrast coding and interpreration with brm package

Interfaces brms
1

I was wondering if my semantic was correct when I constructed the orthogonal contrasts.

I have 3 factors: bodysite, Pattern and Strength. I did the following coding

Bodysite Contracts

contrasts(df.cleaned$Bodysite) <- cbind(back_vs_front=c(-1, -1 ,2), chest_vs_abdomen=c(-1, 1, 0))

Pattern Contrasts

contrasts(df.cleaned$Pattern) <- cbind(vertical_vs_horizontal=c(-1, 0 ,1))

Strength Contrasts

contrasts(df.cleaned$Strength) <- cbind(Strong_exhale_vs_Strong_inhale=c(1, -1))

Then I fit the model:

fit.brm.PA.NA = brm(formula = PA - NA. ~ Bodysite + Pattern * Strength + (1 |id),
data = df.cleaned,
file = “PA.NA4.brm”)

then I mutated the four hypothesis:

fit.brm.PA.NA %>%
posterior_samples() %>%
clean_names() %>%
select(starts_with(“b_”)) %>%
mutate(hypothesis_1 = b_intercept + b_bodysiteback_vs_front > b_intercept,
hypothesis_2 = b_intercept + b_bodysitechest_vs_abdomen > b_intercept,
hypothesis_3 = b_intercept + b_patternvertical_vs_horizontal > b_intercept,
hypothesis_4 = b_strength_strong_exhale_vs_strong_inhale + b_intercept > b_intercept,
hypotheses_joint = hypothesis_1 * hypothesis_2 * hypothesis_3 * hypothesis_4
) %>%
summarize(hyp1 = sum(hypothesis_1)/n(),
hyp2 = sum(hypothesis_2)/n(),
hyp3 = sum(hypothesis_3)/n(),
hyp4 = sum(hypothesis_4)/n(),
hyp_joint = sum(hypotheses_joint)/n())

Is this correct?

Please also provide the following information in addition to your question:

  • Operating System:
  • brms Version:
2

Hi, I am sorry I didn’t respond earlier. I usually don’t work with orthogonal contrasts, but since the way contrasts work is independent of the particular implementation (i.e., brms, lme4, etc.) the contrasts should look the same for all implementations. In other words, if this formulation is correct for other software, it is very likely correct for brms.

1 Like