Alter-nomination event data for Relational Hyperevent Models of personal network elicitation

This dataset accompanies the manuscript "Modeling the elicitation of alters in personal network studies". It contains the event stream used to estimate the Relational Hyperevent Models (RHEM) reported in the paper. Data were collected in a rural Romanian community over four waves (September 2023; March–April 2024; March–April 2025; September 2025) as part of the Horizon Europe 4P-CAN project. A total of 142 egos contributed [353/355] alter-set nomination events. The file events.csv is in long format: each row is one record with the columns e.id (record identifier), source and target (integer actor identifiers in a shared ego-alter ID space), time (an ordinal counter encoding event order, not calendar time), type (the kind of record), and weight. The type column distinguishes: alter_nomination (an ego nominates an alter); aatie (a reported alter-alter tie); link_tracing (one actor referred another as a potential respondent); family and household (group co-membership); event (co-participation in one of five community events); and the actor-level covariates female, age, bmi, procfood, is.ego.w1 - is.ego.w4, edu.w1 - edu.w4, and smoking.w1 - smoking.w4. Covariate and group-membership records are encoded as self-loops (source == target), with the actor's value carried in weight; missing covariate values are coded as weight = -1; non-integer covariate weights arise from averaging over non-missing values. The alter_nomination records aggregate into the [353/355] alter-set nomination events (one event per ego per wave). Actor identifiers are anonymized integers and cannot be linked to individuals. Full covariate definitions are given in Section SI.1 of the manuscript's supporting information. Keywords: personal networks; relational hyperevent models; RHEM; alter elicitation; alter-alter dependence; name generator; egocentric networks; 4P-CAN project