Cragg and Uhler (1970) [101] In an econometric article on the car purchasing behaviour, propose (in a footnote) R2LR for multinomial logistic models.
Maddala (1983) [252] In a book on models for qualitative responses, discusses R2LR and a R2 for grouped (binomial) data, analogous for a generalised R2 for weighted least squares.
Aldrich and Nelson (1984) [16] Propose R2=L2/(N+L2).
Magee (1990) [254] Makes a connection between test statistics for the overall hypohesis `predictors related to the response' and R2 in linear models, i.e. R2=Wald/(N+Wald) = 1-exp(-LR/N) = LM/N (the last is only mentioned in passing). Relates this to existing pseudo R2 measures for other models.