# Supplementary R code for: # Insightful spiders? Pholcus phalangioides solve a prey capture problem differently across repeated trials #Madison A. Rittinger*, Ava Mueller, and Rafael L. Rodr?guez #Behavioral & Molecular Ecology Group, Department of Biological Sciences, University of Wisconsin - Milwaukee, Wisconsin, USA. #* Corresponding author; email: ritting2@uwm.edu; ORCID ID: 0000-0001-6326-1572 #running the packages for the below analyses library("readxl") library("glmmTMB") library("car") library("multcomp") library("writexl") library("tidyverse") library("lme4") library("tidyr") #Importing data for assessing difficulty of the prey capture problems models ################## #these data include all 3 treatments and trials in which the spiders failed or successfully solved the problem SFdata<-read_xlsx("Problemsolvingms_data.xlsx_FINAL", na= "na", sheet=1) #correcting the variable types for the following SFdata$treat<-as.factor(SFdata$treat) SFdata$trial<-as.ordered(SFdata$trial) SFdata$success<-as.factor(SFdata$success) #sub-setting data so it only includes the spider's first trial difficultdata<-subset(SFdata, success==1| reason=="spider") difficultdata<-subset(difficultdata, trial==1) #subsetting data to only include data from females to ensure males did not interfere with our results femearly<-subset(difficultdata, sex=="F") #Assessing difficulty of the prey capture problems############### #logistic regression with whether the spider solved the treatment problem as the binomial response variable failmod <- glmmTMB(data=difficultdata, success ~ treat+sex+tibia, family=binomial (link="logit")) Anova(failmod, test="Chisq", type="III") #testing if removing males interfered with main result femfailmod <- glmmTMB(data=femearly, success ~ treat+tibia, family=binomial (link="logit")) Anova(femfailmod, test="Chisq", type="III") #post hoc tests between the treatments to determine which were significantly different using Tukey adjustment summary(glht(failmod, mcp(treat="Tukey"))) #visually check residuals res <- resid(failmod) plot(fitted(failmod), res) qqnorm(res) #Importing data for assessing qualitative and quantitative variation in solutions############## #these data include trials in which the spider caught the designated number of flies d<-read_xlsx("Problemsolvingms_data_FINAL.xlsx", na= "na", sheet=2) #correcting the variable types for the following d$treat<-as.factor(d$treat) d$success_number<-as.ordered(d$success_number) d$solution<-as.factor(d$solution) d$perc_caught<-as.factor(d$perc_caught) d$rough_prey_dist<-as.factor(d$rough_prey_dist) #subsetting data to only include trials for the multiple fly treatments sol_data<-subset(d, solution=="i"| solution=="pb"| solution=="g") #subsetting data to only include data from females to ensure males did not interfere with our results fem<-subset(sol_data, sex =="F") #Assessing qualitative variation in solutions######################################## #logistic regression with the type of solution used as the binomial response variable slnmod <- glmmTMB(data=sol_data, solution ~ success_number+tibia+flies_caught+rough_prey_dist+ (1|id), family=binomial (link="logit")) Anova(slnmod, type=3) #testing if removing males interfered with main result Fslnmod <- glmmTMB(data=fem, solution ~ success_number+flies_caught+rough_prey_dist+tibia+ (1|id), family=binomial (link="logit")) Anova(Fslnmod, type=3) #testing if ID matters slnmod2 <- glmmTMB(data=sol_data, solution ~ success_number+tibia+flies_caught+rough_prey_dist, family=binomial (link="logit")) anova(slnmod,slnmod2)#ID not sig #testing if removing males interfered with main result Fslnmod2 <- glmmTMB(data=fem, solution ~ success_number+flies_caught+rough_prey_dist+tibia, family=binomial (link="logit")) anova(Fslnmod, Fslnmod2) #visually check residuals res <- resid(slnmod) plot(fitted(slnmod), res) qqnorm(res) #Which solution is caught the most flies############### #logistic regression with the percentage of prey caught as the binomial response variable. #Percentage in this context only includes two possible outcomes: spiders could catch 67% of flies (two of three) or 100% of flies (two of two or three of three) percmod <- glmmTMB(data=sol_data, perc_caught ~ solution+ (1|id), family=binomial (link="logit")) Anova(percmod, type=3) #testing if ID matters percmod2 <- glmmTMB(data=sol_data, perc_caught ~ solution, family=binomial (link="logit")) anova(percmod,percmod2)#ID is sig #Does what happened previously affect next solution? ###################################### #subsetting trials in which the spider succeeded either the first or second time only onevtwo_data<-subset(sol_data, success_number<=2) #reformating data for ease of analysis onevtwo_wide<-pivot_wider(data=onevtwo_data, names_from = success_number, id_cols=c(id,web_cover,sex,tibia), values_from = c(solution,perc_caught,temp,endtime)) #logistic regressions with the type of solution spiders used in the second trial as the response variable sln2_mod<-glmmTMB(data=onevtwo_wide, solution_2~endtime_1+perc_caught_1, family=binomial (link="logit")) Anova(sln2_mod, type=3) sln2_mod2<-glmmTMB(data=onevtwo_wide, solution_2~solution_1, family=binomial (link="logit")) Anova(sln2_mod2) #Assessing quantitative variation in solutions################ #testing assumption of normality shapiro.test(sol_data$endtime)#not normally distributed shapiro.test(sqrt(sol_data$endtime))#normally distributed #adding transformed data back to df sol_data$sqrtendtime<-sqrt(sol_data$endtime) fem$sqrtendtime<-sqrt(fem$endtime) #linear mixed model with the time until the solution as the continuous response variable timemod <- lmer(data=sol_data, sqrtendtime ~ success_number+solution+tibia+temp+ (1|id)) Anova(timemod, type="3", test.statistic = "F") #testing if removing males interfered with main result Ftimemod <- lmer(data=fem, sqrtendtime ~ success_number+tibia+temp+solution+ (1|id)) Anova(Ftimemod, type="3", test.statistic = "F") #testing if ID matters timemod2 <- lm(data=sol_data, sqrtendtime ~ success_number+tibia+temp+solution) anova(timemod,timemod2)#ID not sig #visually check residuals res <- resid(timemod) plot(fitted(timemod), res) qqnorm(res) #Testing for trial and error######################### #generalized linear model with a Poisson distribution and the number of failed trials before the spider solved the problem as the continuous response variable TE_mod_all<-glmmTMB(data=d, spfails_before ~ success_number+treat+sex+web_cover+(1|id), family=poisson (link="log"), ziformula = ~1) Anova(TE_mod_all, type=3) #testing if removing males interfered with main result fem_TE_mod<-glmmTMB(data=fem, spfails_before ~ treat+success_number+web_cover+(1|id), family=poisson (link="log"), ziformula = ~ 1) Anova(fem_TE_mod, type=3) #post hoc tests between the treatments to determine which treatment spiders failed more in summary(glht(TE_mod_all, mcp(treat="Tukey"))) #testing if ID matters TE_mod2<-glmmTMB(data=d, spfails_before ~ success_number+treat+sex+web_cover, family=poisson (link="log"), ziformula = ~1) anova(TE_mod_all,TE_mod2)#ID is sig #visually check residuals res <- resid(TE_mod_all) plot(fitted(TE_mod_all), res) qqnorm(res)