Logistic model - numerical vs. categorical covariates & target

#1
I am using a powerful analytics tool built by Oracle (Data Miner). I am trying to determine the probability of death for a patient given the presence of certain diagnosis codes. There will be up to 40 possible codes as column references (covariates) in the model.
Any reason why I must convert the presence indicator to a numeric 0/1 design variable ? Same thing with the target variable (death) - must it be 0/1 (0=alive, 1=death) ?

I noticed that SAS users do this regularly, but my modelling package allows for categorical "Y","N" instead.
 

noetsi

Fortran must die
#3
I am not sure I understand what you are asking, but if you have a choice between coding the DV as a percentage (and thus quantiative) variable or 1/0 I would chose the former. With the former you can use (assuming other assumptions are met) linear regression which has a number of advantages over doing logistic regression (what you have to do with 1/0 coding).
 
#4
I am not sure I understand what you are asking, but if you have a choice between coding the DV as a percentage (and thus quantiative) variable or 1/0 I would chose the former. With the former you can use (assuming other assumptions are met) linear regression which has a number of advantages over doing logistic regression (what you have to do with 1/0 coding).
In my case, it's just logistic - True / False for both target variable and dependent covariates.
But that makes sense....and could be the reason for using the 1/0....so that linear regression would be a possibility.