Andy Grogan-Kaylor

Multilevel Thinking Multilevel models have become very common in social research, however, there are aspects of these models that are explored only infrequently. Those that rarely make their way into courses and training...

Now available both in print and online: books.google.com/books/about/...

Why care about these technicalities?

Interactions are one of our major statistical tools for examining similarity or difference in statistical relationships across groups, identities, or contexts.

Interactions in logistic regression can be particularly challenging to work with: agrogan1.github.io/newstuff/cat...

Interactions And Main Effects Need To Be Interpreted Together – Andrew Grogan-Kaylor A Mathematical Perspective

Both main effects and interactions need to be interpreted together:
agrogan1.github.io/posts/intera...
agrogan1.github.io/posts/intera...

Interactions Are Symmetric – Andrew Grogan-Kaylor

Interactions are often thought of as encompassing a main effect (x), and a moderator (m), but because of the commutative property of multiplication, interactions are essentially symmetric: agrogan1.github.io/posts/intera...

Interactions and Moderation – Andrew Grogan-Kaylor

For many models, adding a main effect and interaction for group can be seen as giving each group its own regression line with its own intercept and its own slope:

agrogan1.github.io/posts/intera...
agrogan1.github.io/posts/intera...

slope of a mountain

A thread on #interactions and #moderation.

How Does Context Shape Caregivers' Psychosocial Stimulation? A Study of Georgia and Mongolia

arcg.is/1jHXPX1

I argue that multilevel modeling is a principled way to explore human variation and commonality.

I try to provide a tutorial for applied researchers, and to bring "down to earth" some of the more advanced ideas about multilevel modeling that I see in the methodological literature, but rarely see in applied work.

Multilevel Thinking: Discovering Variation, Universals, and Particulars in Cross-Cultural Research Abstract. This book introduces the ideas of multilevel modeling. Chapters cover: the cross-sectional model; the longitudinal model; models for binary outco

My new book, Multilevel Thinking, is now published in Electronic version by Oxford University Press: academic.oup.com/book/60530. #stata #rstats #julialang

Violence between parents or in the community increases risk of child punishment Exposure to higher rates of violence in the larger social context may spill over to family violence, including caregivers' use of physical child abuse, according to a new University of Michigan study.

UM News release about our research: news.umich.edu/violence-bet.... See also: globalfamilies.quarto.pub/spillover-of...

DOI for the Open Access article: doi.org/10.1016/j.ch...

Just similarity and difference across groups or countries, which I think, multilevel models let us start to see. Simply the size of the random slopes. Now that I’ve written the book, I need to figure out even better ways to quantify this.

I’d like to share that my book, Multilevel Thinking: Discovering Variation, Universals, and Particulars in Cross-Cultural Research is available for pre-order at Oxford University Press: global.oup.com/academic/pro...

* Includes appendices for estimating MLMs in #Stata, #Rstats, and #Julialang.

Countries in UNICEF MICS Data

We've published a new paper: Spillover of Macro-Level Violence to Parental Physical Abuse of Children in Low- and Middle-Income Countries. The paper should be online shortly, but in the meantime, we've prepared a scrollable summary here: globalfamilies.quarto.pub/spillover-of...

Thank you for the kind words!

Orphan Discipline and Child Neglect An Analysis from 48 Countries

ArcGIS StoryMap visualization of our new article on parenting of orphans: arcg.is/1HjKyW2

treemap showing that most of the world's children are not protected from physical punishment

Same information as a treemap.

Most of the world's children are not protected from corporal punishment.

Cartogram showing that most of the world's children are not protected from #corporal_punishment #rstats #gis

One model accounts for group, while the other does not.

“Be Less Wrong”, the importance of multivariate, quantitative thinking, especially when thinking about treatments, programs, and interventions for social issues: globalfamilies.quarto.pub/be-less-wrong/
#closeread #quarto

I experimented with using the new #closeread (closeread.dev/gallery/) extension for #Quarto to share the results of an article we published last year on parenting and child development in 60 countries: globalfamilies.quarto.pub/parenting/