@MatthewDrury thanks for the response. Have you considered that making features in this context is not part of the learning algorithm - i.e. the label or dependent variable has nothing to do with the making of the features.

@user5552894 (@MatthewDrury) if you have a substantive question about text mining, please ask it on the main site. That isn't a chat item.

@gung I was not involved in that suspension but I have investigated the circumstances and strongly agree with it. I cannot speak about this particular circumstance in public.

I would like to say, purely as a general proposition, that any user who repeatedly badgers new users and directs public derogatory remarks at other users will be suspended briefly to wait for emotions to cool. If such behavior were to continue after such a suspension, a longer suspension would ensue.

@whuber, I wondered if there was more going on than I could see. Thanks for letting me know.

Sorry for the inappropriate question and for my bad English: what is the name of this kind of distribution (overlapping). For example: we want to estimate the distribution of the population's age. Men follows F1(mean=50,sd=5), women follows F2(mean=70,sd=10). What's it's called the joint distribution of men and women?

like in this graph: rasch.org/rmt/gifs/101over.gif

@Matemattica "Joint distribution" does not apply. Perhaps you are asking about a mixture distribution?

Once again sorry for my bad English, I know that the term joint distribution isn't appropriate but I can not find the exact word.

OK I found it by accident, bimodal distribution is what I was refering to :) thanks!

@Matemattica "Bimodal" might not mean what you intended. For instance, if you are talking about equal numbers of men and women and you replace "70" by "60", the distribution of their heights will be unimodal.