Statistical Analysis - Details

The published manuscript contains important details regarding statistical analyses.

• The “Statistical analyses” methods section contains a narrative explanation and justification for analyses.
• Table S3 lists the analysis type, data and variables included, and results from the final model for each dependent variable.

Below we provide general information about the statistical software used, and specific information for each individual analysis type.

Statistical Software

• We used the program NCSS (v11.0.24) to conduct all statistical analyses except quantitative genetic analyses (see below).
• A free trial of the newest version of this proprietary software (NCSS 2024) is available at https://www.ncss.com/download/ncss/free-trial/.
• This software is based upon a graphical user interface, and does not involve user-generated code.
• Our settings were based on recommendations from the NCSS Documentation Manual, which is available at https://www.ncss.com/software/ncss/ncss-documentation/.
• To ensure clarity and reproducability within this GUI-based program, we have itemized all settings for each analysis type below.

Mixed Models
Analyses were completed using the “Mixed Models-Repeated Measures” procedure.

• Subject is included as a random effect
• Model includes random intercept and uses Type III sums of squares
• Calculation of test statistics (F, df, p) are based upon methods described in Kenward and Roger (1997; https://doi.org/10.2307/2533558)
• Variables Tab
  • Response: Variable depending upon analysis (see Table S3).
  • Subjects: Table S3 in the manuscript lists the subject for each analysis. Most use the female (“BFID”) as the subject. Exceptions include Nestling Feeding in which the pair (“Group”) was the subject, and relative breeding onset day in which site/year (“SiteFirstEggDay”) was teh subject.
  • Times: Left blank, which causes time values to be assigned sequentially. Because spreadsheets are sorted chronologically this results in accurate order assignment.
  • Between and Within Fixed Factors: Variable depending upon analysis (see Table S3).
  • Covariates: Variable depending upon analysis (see Table S3).
  • Pattern: “Diagonal:Random” was the starting point for all analyses. This is similar to the default “diagonal” repeated component pattern, but it adds a random subject term. The random subject term was removed if the random component parameter estimate was zero or if model simplification was necessary for proper convergence.
  • Force Positive Correlations: Not selected
  • Model (Fixed Terms): Includes all fixed factors, covariates, and interactions (see Table S3).
  • Intercept: Selected to include the intercept in the model. The only exception was the analysis of PC2, which properly converged only after simplifying the model by removing the intercept.
• Maximization Tab
  • Likelihood Type: Restricted Maximum Likelihood (REML)
  • Solution Method: Newton-Raphson
  • Write Solution To: Left blank
  • Max Fisher Scoring Iterations = 5
  • Max Newton-Raphson Iterations = 40
  • Lambda = 1
  • Convergence Criterion = 1E-6
  • Zero (Algorithm Rounding) = 1E-12
  • Variance Zero = 1E-6
  • Correlation Zero = 1E-6
  • Max Retries = 10
• Reports Tab
  • Alpha = 0.05
  • All remaining options: Variable depending upon need

Multiple Regression
Analyses were completed using the “Multiple Regression” procedure.

• Variables, Model Tab
  • Y: Variable depending upon analysis (see Table S3).
  • Numeric X’s: Variable depending upon analysis (see Table S3).
  • Categorical X’s: Variable depending upon analysis (see Table S3).
  • Default Recording Scheme: Binary
  • Default Reference Value: First After Sorting
  • Weights: Left blank
  • Regression Model: Includes numeric and categorical X’s (see Table S3).
• Reports Tab
  • Tests Alpha: 0.050
  • Assumptions Alpha: 0.20
  • Confidence Level: 95%
  • Select Reports: Variable depending upon need.
• Report Options Tab
  • Variable depending upon need.

Logistic Regression
Analyses were completed using the “Logistic Regression” procedure.

• Variables, Model Tab
  • Y: Variable depending upon analysis (see Table S3).
  • Reference Value: Left blank
  • Numeric X’s: Variable depending upon analysis (see Table S3).
  • Default Recording Scheme: Binary
  • Default Reference Value: First After Sorting
  • Frequencies: Left blank
  • Validation Filter: Left blank
  • Regression Model: Includes numeric and categorical X’s (see Table S3).
  • Remove Intercept: Left blank
  • Priors: Equal across Y Values
• Subset Selection Tab
  • Search for the best subset from the X’s: Left blank
• Iterations Tab
  • Maximum Iterations: 20
  • Iteration Termination: 0.000001
• Reports Tab
  • Variable depending upon need.
• Report Options Tab
  • Confidence Level: 95%
  • All remaining options: Variable depending upon need

Analysis of Variance
Analyses were completed using the “General Linear Models (GLM)” procedure.

• Variables Tab
  • Response Variable: Year2RS
  • Covariate Variable(s): None
  • Factor Variables: Son plumage, Site, Year
  • Factor Types: Fixed
  • Factor Comparisons: None
  • Which Model Terms: Up to 1-Way
• Reports Tab
  • Test Alpha: 0.05
  • All remaining options: Variable depending upon need

Principle Components Analysis
Analyses were completed using the “Principle Components Analysis” procedure.

• Spreadsheet does not contain records with missing data for variables included in PCA analysis (missing data are omitted).
• This procedure scales and centers data for each variable and presents factor scores that are scaled and centered.
• Variables Tab
  • Variables: NestingAttempts, EggsTotal, NestlingsTotal, FledgedTotal
  • Data Input Format: Regular data
  • Robust Covariance Matrix Estimation: Not selected
  • Missing value estimation: None
  • Robust Weight: 4.0
  • Maximum Iterations: 6
  • Matrix Type: Correlation
  • Factor Rotation: None
  • Factor Selection-Method: Percent of Eigenvalues
  • Factor Selection-Value: 100
• Reports Tab
  • Select Reports: Variable depending upon need.
  • Alpha = 0.05
  • Report Options: Variable depending upon need.

Statistical Analysis - Data Files

All data files are provided in CSV format. The first row of each spreadsheet contains the column names. File names indicate the dependent variables analyzed with each data file. Model details and statistical results for each dependent variable can be found in Table S3 of the associated manuscript.

Spreadsheets provided only include data records that were included in analyses. Records have been omitted if they were missing data that were necessary for all associated analyses. Data does not need to be filtered prior to analysis unless stated specifically in the field descriptions for a given spreadsheet.

Spreadsheets may still contain records with missing data for two reasons. First, if the spreadsheet was used for a single analysis where the a priori model did not include the variable with missing data. Second, if the spreadsheet was used for multiple analyses but the variable with missing data was not included in all of them. Missing values are represented as blanks, as recommended in the NCSS manual (see above). If the variable with missing data is included in a statistical model this program automatically omits the associated record from the analysis. Records with missing data are only included in analyses when the variable with missing data is not included in the model.

Detailed field descriptions for each spreadsheet are provided below:

1.#Eggs.FemaleAge.csv
This spreadsheet was used to analyze the number of eggs laid and to test for assortative pairing by age.

2.PCA.PC1.PC2.csv
This spreadsheet was used to conduct a principal components analysis (PCA) of correlated measures of female reproductive effort. It was also used to analyze the resulting principal components.

3.NestlingFeeding.csv
This spreadsheet was used to analyze the relative contribution of each parent to feeding nestlings.

4.FoodPerNestling.PreySize.csv
This spreadsheet was used to analyze the average quantity of food that parents delivered to each nestling and the average size of prey items females fed to their nestlings.

5.NestlingMass.FledgingRate.BreedingOnsetDay.csv
This spreadsheet was used to analyze nestling mass, the percent of nestlings that fledged, and female breeding onset day.

6.RecruitmentRate.GrandOffspring.csv
This spreadsheet was used to analyze the rate of offspring recruitment (measured as survival to year after birth) and the number of grand-offspring produced by sons in the year after their birth.

7.BreedingOnsetDay(Site/Year).csv
This spreadsheet was used to analyze shifts in female breeding onset day in response to average breeding onset on that site in that year. In other words, it analyzed the extent to which females advanced breeding in “early-breeding” years.

8.OffspringPlumage&RS.csv
This spreadsheet was used to analyze the plumage color and reproductive success of first-year males based upon their developmental environment.

9.BodyCondition.csv
This spreadsheet was used to analyze changes in female body condition across the course of the breeding season. Body condition was estimated from fat stores.

10.Survival.csv
This spreadsheet was used to analyze female survival from her first breeding season to the following year. All included females are one year old.

11.Lifespan&LifetimeRS.csv
This spreadsheet was used to analyze female lifespan and lifetime reproductive success.

Quantitative Genetic Analysis - Details

The published manuscript contains important details regarding quantitative genetic analyses.

• The “Quantitative genetic analyses” methods section contains a narrative explanation and justification for the analyses.
• Table S1 displays results from this analysis.

Quantitative Genetic Analysis - Code File

We performed the quantitative genetic analyses in R, and used the provided script file to estimate quantitative genetic variance components for total genetic offspring using MCMCglmm. A description of the code is below:

QGenScript.Final.Sep24.R
This file contains the R script used for quantitative genetic analyses. The code contains annotations that describe the purpose of each set of commands and explain possible errors and their interpretation.

We conducted quantitative genetic analyses in R using the most updated versions of the program and loaded packages available at the time (analyses completed on September 4, 2024):

• R (v4.4.1; https://www.R-project.org)
• MCMCglmm (v2.36; https://doi.org/10.32614/CRAN.package.MCMCglmm)
• remotes (v.2.5.0; https://doi.org/10.32614/CRAN.package.remotes)

Quantitative Genetic Analysis - Data Files

Data files used in quantitative genetic analyses are provided in both RAT and CSV formats. Users should only use one file type when running code (do not load both). RAT files are preferentially used by Program R, and thus the code provided loads the RAT files. However, the code provides annotation that can be used to instead load CSV files if preferred by users.

ID numbers in each file have been modified to indicate birth year. Sequential numbers representing years from 1996 (0) to 2011 (15) were added before the 4-digit bird ID number (e.g. if bird 1389 was born in 2003, that bird would be listed as 71389). Thus, ID number simultaneously identifies the bird and its birth year.

Missing values in the spreadsheets below are represented by “NA”.

Detailed field descriptions for each spreadsheet are provided below:

AllTotalGenOff.csv
This is the primary data file.
This same data is also provided in .Rat format.

DougPed.csv
This data file only contains a pedigree.
This same data is also provided in .Rat format.