Standard deviation of k-th class, corresponding to j-th feature

30 Views Asked by At

I'm trying to calculate Fisher's score for a feature from a data set. The fisher score for the $j$-th feature:

$$F(x^{j}) = \frac{\sum^c_{k=1} n_k(\mu^j_k - \mu^j)^2{}} {\sum^c_{k=1}n_k(\sigma^j_k)^2}$$ where $c$ is the number of different classes, $n_k$ is the size of the $k$-th class, $\mu^j_k$ and $\sigma^j_k$ are the mean and standard deviation of the $k$-th class (respectively), corresponding to the $j$-th feature, $\mu^j$ is the mean deviation of the whole data set corresponding to the $j$-th feature.

It says that $\mu^{j}_{k}$ is the mean deviation of the $k$-th class, corresponding to the $j$-th feature. Standard deviation calculation requires $X$, $\mu$, and $N$, which are value in data set, mean of the data set and the population count respectively. $$\sqrt{\frac{\sum{|x-\mu|^2}}{N}}.$$

I'm a bit confused whether I have to find $\mu$ in the standard deviation formula from the $k$-th class subset or the overall data set and the same with the count—number of data entries from the whole data set or just the subset?

The whole paper on the topic from where I got the formula