Some of the long tail very discrete data, eventually even an order of magnitude higher
Thus, the need to convert the data - to shorten the tail
Standard deviation in order to see, so that it becomes possible normal / normal distribution is very similar, and thus the use of a recursive linear modeling or other means
There are two ways to transform:
- Variables used in the wrapper
qplot(x=log(friend_count),data = pf)
qplot(x=sqrt(friend_count),data = pf)
Disadvantages: the x-axis value will be changed accordingly
- Use scale layer
ggplot(aes(x = friend_count), data = pf) + geom_histogram() + scale_x_log10()
ggplot(aes(x = friend_count), data = pf) + geom_histogram() + scale_x_sqrt()
Advantages: the x-axis value remains the original value