Mamaros and Sacerdote (2006) look at variables determining volume of email. Their study population consists of students and recent graduates of Dartmouth; the study period year is one academic year. Let Yij be the number of emails exchanged between person i and person j , while Xi is a 1×p vector describing characteristics of person i, and β is a p×1 parameter vector. Furthermore, Xij is a 1×q vector describing characteristics of the pair (i, j ), and γ is a q×1 parameter vector. Let exp(x) = ex , p(y|λ) = exp(−λ)λy /y!.
To estimate parameters, the authors maximize 
1≤iYij #
# exp (Xiβ + Xjβ + Xij γ )
as a function of β and γ , where n is the number of subjects. (Some details are omitted.) Comment briefly on the data analysis.