statisticalrethinkin..

282 11. COUNTING AND CLASSIFICATIONR code11.3Estimate S.E. 2.5% 97.5%a 0.32 0.09 0.14 0.5is implies a MAP probability of pulling the le lever of logistic(0.32)≈ 0.58, with a95% quadratic estimate interval of 0.54 to 0.62:logistic( c(0.14,0.5) )[1] 0.5349429 0.6224593So we have to note at the start that the chimpanzees, on average, exhibit a preference for thelehand lever.Now fit the next two models, which allow us to investigate the hypothesis. We don’t fita model with only condition, because we don’t think there is any reason to suspect thatchimpanzees pull the lehand lever more or less when another chimpanzee is present. Incontrast, we have reason to suspect that they pull the lehand lever more when the lehandlever is both 1/1 option and another chimpanzee is present. is means we want to estimatethese additional two models:L i ∼ Binomial(p i , 1)p ilog = α + β P P i1 − p iα ∼ Normal(0, 10)β P ∼ Normal(0, 10)R code11.4L i ∼ Binomial(p i , 1)p ilog = α + (β P + β PC C i )P i1 − p iα ∼ Normal(0, 10)β P ∼ Normal(0, 10)β PC ∼ Normal(0, 10)where P is prosoc.left, a 1/0 variable indicating when the 1/1 option was on the le side.C is condition, a 1/0 variable indicating when another animal was present. Note that thesecond model above has a two-way interaction, but it doesn’t consider one of the main effects.is is because we are assuming that there is no direct effect of condition on pulling le versusright. You may want to fit a model that considers the main effect of condition, at the end ofthis section, so you can see if inferences change any.To fit these models, just build the linear model of log-odds:m10.2