Identifying the factors that influence organism growth is crucial to understanding how conditions in the wild affect ecological outcomes over generations. The aim of this study is to understand conditions that influence the larval growth rates and adult weights of Pieris rapae. In the data we used, mothers reared on one of two different plant genuses, Barbarea and Berteroa, and oviposited on one of those same two plants. Response metrics in offspring development were recorded.
We used ANOVA to analyze the extent to which the maternal host (the plant on which the mothers reared) and the larval host (the plant on which the mothers oviposited) affect larval growth rates and adult weights of P. rapae individuals.
Statistical Analysis
Two two-way ANOVA analyses were performed to investigate the relationship between the two factors (maternal host and larval host). A linear model was fitted to each response variable and both factors, producing one model for the effect of the factors on growth rate and one model for the effect of the factors on adult weight.
Results
We found that mean adult weights were higher when Barbarea was the larval host. For individuals with Barbarea larval host, their mean adult weights were 65.4 and 66.7 mg for Barbarea and Berteroa as maternal host, respectively. Meanwhile, individuals with Berteroa as larval host had mean adult weights of 53.5 and 50.4 mg for Barbarea and Berteroa as maternal host, respectively (Figure 1).
As for growth rates, we observed a similar pattern, where individuals with Barbarea as larval host had the highest mean growth rates, at 0.084 and 0.078 mg/day for Barbarea and Berteroa as maternal host, respectively. Individuals with Berteroa as larval host had mean growth rates of 0.064 and 0.055 mg/day for Barbarea and Berteroa as maternal host, respectively (Figure 1).
Figure 1. For each response variable, means were higher for individuals in the Barbarea larval host group than for the Berteroa larval host group. Additionally, within maternal host groups, mean response was consistently higher for individuals within the Barbarea larval host group than the Berteroa larval host group, with the exeption of mean adult weight of individuals with Barbarea maternal host, where individuals with Berteroa larval host had the higher mean adult weight.
We found the effect of larval host on adult weight to be significant (p<0.001), while the effect of maternal host on adult weight was not significant (p=0.77). The interaction between the two factors was represented by p=0.054. Furthermore, we found the effect of larval host, maternal host, and the interaction of larval and maternal host on growth rate to be significant (p<0.001, p<0.001, p=0.016, respectively).
Conclusions
Following the statistical analysis presented above, it is clear that the plant a female P. rapae individual choses to oviposit on has a sizable impact on her offspring’s growth rate, and ultimately their adult weight. Statistical analysis using ANOVA showed that it is not only the above effect that has a statistically significant impact on the responses measured, but also the effect of maternal host and the effect of the interaction of larval and maternal host on growth rates.
The significance of the effect of the interaction of larval and maternal host on adult weight was not below alpha, however, it was quite close at p=0.054. While this cannot be said to be a statistically significant observation, it is obvious that there is quite some interaction happening, and it is indeed possible that the interaction is biologically important, even if a significant p-value was not observed in this case.
Code
The code below is a copy of the code in Script/Analysis.R. You may run the aforementioned file.
── Conflicts ────────────────────────────────────────── tidyverse_conflicts() ──
✖ dplyr::filter() masks stats::filter()
✖ dplyr::lag() masks stats::lag()
ℹ Use the conflicted package (<http://conflicted.r-lib.org/>) to force all conflicts to become errors
library(svglite)# Import datadf <-read.csv(here("Data", "butterflies.csv"))# Convert to tibbledata <-as_tibble(df)# Select relevant datadata <- data %>%select(LarvalHost, MaternalHost, AdultWeight, GrowthRate)# Fit linear models and perform ANOVAm_gr <-lm(data$GrowthRate ~ data$LarvalHost * data$MaternalHost -1)m_aw <-lm(data$AdultWeight ~ data$LarvalHost * data$MaternalHost -1)anova_gr <-anova(m_gr)anova_aw <-anova(m_aw)# Save ANOVA output to csvwrite.csv(anova_aw, here("Output", "anova_aw.csv"))write.csv(anova_gr, here("Output", "anova_gr.csv"))# Pivot data along responses to help with plotting facetsdata_p <- data %>%pivot_longer(cols = GrowthRate:AdultWeight, names_to ="Response", values_to ="Value" )# Create dataframe data_summary to contain means of boxes in plot and savedata_summary <- data_p %>%group_by(LarvalHost, MaternalHost, Response) %>%summarize(mean_response_value =mean(Value))
`summarise()` has grouped output by 'LarvalHost', 'MaternalHost'. You can
override using the `.groups` argument.
write.csv(data_summary, here("Output", "data_summary.csv"))# Create custom labels for the plot facetsresponse.labs <-c("Growth Rate (mg/day)", "Adult Weight (mg)")names(response.labs) <-c("GrowthRate", "AdultWeight")# Plot and savedata_p %>%ggplot(aes(x =factor(MaternalHost, level =c("Barbarea", "Berteroa")), y = Value,fill = LarvalHost )) +geom_boxplot() +labs(x ="Maternal Host", y =NULL, fill ="Larval Host") +facet_wrap(facets =vars(Response), nrow =2,strip.position ="left", # Move facet labels left to act as y-axis labelsscales ="free_y", labeller =labeller(Response = response.labs) ) +theme(strip.background =element_blank(), # Clear background behind facet labelsstrip.placement ="outside"# Place labels outside of axis ticks )
ggsave(here("Output", "Fig1.svg"), width =1100, height =1100, units ="px")
