ManyBabies

Re-analysis of data from the ManyBabies1: Infant-directed Speech Preference project.

• For more information about the ManyBabies project, see http://manybabies.stanford.edu/
• The data and main analysis can be found at https://github.com/manybabies/mb1-analysis-public, stimuli, protocol, and further documentation is at https://osf.io/re95x/
  • This is the paper (which we ask to cite if you do anything with this dataset) The ManyBabies Consortium. (2019). "Quantifying sources of variability in infancy research using the infant-directed speech preference." In press at Advances in Methods and Practices in Psychological Science (AMPPS). Preprint

The repository contains:

• MB1_analysis.jmd - The main script, which

  1. Reads in ManyBabies 1 data
  2. Shapes it as needed
  3. Reproduces the main analysis from the paper at https://github.com/manybabies/mb1-analysis-public
  4. Fits the preregistered maximal model
  5. Simplifies the random effects by inspecting the output of rePCA and the variance components
  6. Shows what can go wrong with multi-lab data (subid vs subid_unique)
  7. Re-processes more messy data to apply different cleaning criteria to deal with censoring (which you can adjust)
  8. Re-runs the analysis

• MB1_analysis.ipynb - The corresponding Jupyter notebook that you can run in your browser, but which differs from the converted version of the .jmd (split code blocks to see all output, converted R code cell)

• MB1_minimal_lmer.R - The R code needed to reproduce the main analysis of the paper, extracted from https://github.com/manybabies/mb1-analysis-public/blob/master/paper/mb1-paper.Rmd

• intendend_complex_LMM.txt - Output for the preregistered model (which still takes some time to fit)

• Project.toml and Manifest.toml - See https://repsychling.github.io/pkg.html

For instructions how to run code in .jmd and .ipynb files, see https://repsychling.github.io/intro.html

Acknowledgements

This work was supported by the Center for Interdisciplinary Research, Bielefeld (ZiF) Cooperation Group "Statistical models for psychological and linguistic data".