This Is AuburnAUrora

Data on morphology, hatch, survival, and deformities used for SAS modelling and final VC results table


Myers, Jaelen N.
Butts, Ian A.E.
Dunham, Rex A.
Chatakondi, Nagaraj


The dataset includes all the combined data for larval morphology, embryonic survival, hatch success, and rate of deformities. One value is given for each trait and for each basket, in which the three replicate baskets per family were averaged for a total of 120 rows per dataset. All data was analyzed using SAS statistical analysis software (v.9.1; SAS Institute Inc., Cary, NC, USA). Residuals were evaluated for normality (Shapiro–Wilk test) and homoscedasticity (plot of residuals) to ensure they met model assumptions. Data were log(10) or arcsine square root transformed to meet these assumptions when necessary. Alpha was set at 0.05 for testing main effects and interactions. Traits were analyzed independently at each sampling point. Least squared means (LSMs) and VCs (% of overall variation due to random effects) were constructed using the restricted maximum likelihood (REML) method with the SAS PROC MIXED statement. The factorial mating design measures the variance of the male (sire) and female (dam) effects and the sire x dam interaction effects, allowing us to infer the VCs in terms of combining ability as an additional interpretation of maternal/paternal random effects. For variation across temperature: Each trait was analyzed using a multi-factorial ANOVA (with fixed effect of temperature and random effects including maternal/paternal effects and all interactions terms) based on the means for each basket. Denominator degrees of freedom for all F-tests were approximated using the Kenward Roger procedure. A posteriori analyses performed on fixed effects were constructed using Tukey’s multiple comparisons method. To test for significant variability among VCs greater than zero in the PROC MIXED model, likelihood ratio statistics were generated from the -2[Res]tricted log-likelihood estimate of the full model and then with each VC held to 0 using the PARMS statement. The probabilities were halved to account for the one-tailed probability and obtain the significance level for each VC. For variation within temperatures: Separate PROC GLM models were used to analyze the following random effects at each temperature: maternal, paternal, and maternal × paternal interactions. VCs were generated directly for the model using the PROC VARCOMP statement. The following document shows the results of the SAS procedures and the variance components. VCs were calculated as percents and determined to be significant by the p-values. All significant VCs are highlighted in green for each trait.