740 



Fishery Bulletin 88(4), 1990 



and the odds for injured crabs (0.717). Alternatively, 

 the log(ODDS) can be predicted as a linear function of 

 each coefficient by substituting these values into Equa- 

 tion 5, e.g., for active, uninjured crabs, 



ln(f,/f,„) = i^(l„ + l„.v + lin) 



= 2(0.368 + 0.77.5 + 0.167) 

 = 2.619 

 ODDS = exp(2.619) = 1.S.74. 



Thus a 'good' vitality code of 1 was a better predictor 

 of future survival odds than lack of injuries. 



Table 5A shows the observed frequencies, percent 

 survival, and survival odds, as well as the survival odds 

 predicted by the logit model, for each combination of 

 factors. Expected odds were calculated as the product 

 of odds for each combination of interaction represented 

 in Table 4. For instance, the expected survival odds for 

 active (VIT = 1), injured (INJ = 2) crabs equalled 7.05, 

 the i)rciduct of odds for overall survival (2.087), active 



