Divide estimates from contrasts

[TommasoGhilardi]

As we already decided we should change the estimate page to be two separate things

[FrancescPoli]

I am going through all stats pages and will make small edits, I will keep track of them here as an update of estimate pred/mean/contrasts is the main update. I reorganised slighly the tutorial on Linear Models, connecting it more to initial hypothesis (Looking time should differ across conditions). Let me know what you think

[FrancescPoli]

I went through mixed effects tutorial. Again, no big changes, but I kept linking back to our hypotheses. I made some key changes. For example, we were talking about random intercept as capturing "baseline performance", but in this models the dependent variable is looking time, not performance. so I changed that. For random slopes, we talked about "rate of improvement" but again, decrease in looking time is not necessarily an improvement. so I just changed it to the more general "rate of change". I think the tutorial is still quite dense, and the example is not super clear: we first and foremost cared about a difference between simple vs complex condition, and now we are spending lots of time modelling trial number...

I also removed simple linear model from model comparison, as it is technically a wrong model, and so it shouldn't even reach the point of getting compared with other ones...

[FrancescPoli]

I also went through the ModelEstimates tutorial. I think it's a bit simpler but it could improve even further by splitting it in two. Let me know what you think about all this - it would be nice to have the new ones up for BCCCD, but there's still a month left, so no rush!