@sunyam "The removed dummy then becomes the base category against which the other categories are compared." source

So, when you drop the column 'male', the odds for a female or child passenger to survive is compared to the male passenger, and so, if the coeff of 'child' is -2.14, it means child passengers are less likely to survive compared to the male passengers, in other words, it means the odds for a child passengers to survive is less than the odds for the male passengers.

So, Which level you decide to make the base category depends on your goal. In general, If you're just making a prediction, it doesn't really matter. But, you still can drop different columns, and test your prediction accuracy.

这个朋友问的问题，也一直是我的疑惑，由于实际经验少，我还得多多思考这个。还是有点不是太理解。

难道删除一个变量，就是将他变成一种基准吗？不是很理解。