How can we change gender preference into dichotomous variable when the categories are male, female or any in order to conduct logistics regression? How can we change gender into dichotomous variable when the categories are male, female or any in order to conduct logistics regression? 
My research is about gender preferences at the time of birth, and I categories gender preference variable as male, female, or any. 
How can we convert it into two categories? How can I conduct logistics analysis?
 A: Since any would encompass either value for the categorical variable, this would still be a single dichotomous indicator variable and the results indicating any gender would be coded as missing values (i.e., as NA).  When conducting logistic regression with this  as one of your explanatory variables, you can employ standard methods to handle missing data (e.g., multiple imputation or Bayesian methods).  If gender is your response (which seems to be the case from the comments) then these data points are effectively useless, since there is no specified response.
A: You could view this situation as an attempt to use an underlying latent variable for gender preference ranging from strongly prefer a girl through neutral to strongly prefer a boy. What you have actually observed is the latent variable cut into three categories. This seems to be the model which underlies ordered categorical regression so I would suggest fitting that model. You attempt to use multinomial logistic is also possible but it does not take account of the ordered nature  of your three categories. I am afraid I do not use SPSS so cannot tell you which command to use but this page is from a usually very helpful site.
