I ran a model with reaction time as my DV and PWI Condition as one of the fixed factors.
I used contr.sum for all fixed factors. I ran the following model to look for differences in reaction time scores across three different IV's (2 levels each).
test=lmer(PWI_RT ~ PWI_Condition * Distractor_Condition *
Distractor_Language + (1 | Subject) + (1 | Picture_Name),
data=data1 ,],)
The output is as follows:
| Estimate | Std. Error | t value (Intercept)
| 3605.485 | 77.181 | 46.71
PWI_Condition1 | 109.131 | 30.595 | 3.57
The output (deviation coded) shows a positive t value of 2.067 for the PWI condition where Noncognates are coded (-1) and cognates are coded (+1). Hence, I interpret this as cognates take significantly longer to name than non-cognates. However, these results do not match with the raw means. The means for Non-cognates are much longer (in terms of reaction times) than for cognates. How can I explain this result?