Types of Distributions:
Discrete Distributions: Probability concentrated on specific values (e.g., coin flips, number of customers).
- Uniform: All outcomes equally likely (e.g., drawing a card from a shuffled deck).
- Bernoulli: Two possible outcomes (e.g., coin flip, true/false).
- Binomial: Sequence of Bernoulli trials (e.g., number of heads in 10 coin flips).
- Poisson: Frequency of events in a fixed interval (e.g., website visits per hour).
Common Continuous Distributions: Probability spread over a continuous range
- Normal (Gaussian): Bell-shaped curve, symmetric, thin tails (e.g., heights, exam scores).
- T: Similar to normal but with fatter tails, used with limited data.
- Chi-squared: Asymmetric, non-negative, used in hypothesis testing.
- Exponential: Models time between events (e.g., website traffic, radioactive decay).
- Logistic: S-shaped curve, used in forecasting and modeling growth.
Practical
What is the distribution of numerical feature values across the samples?
Observation:
Decision:
What is the distribution of categorical features?
g = sns.FacetGrid(df, col='Survived')
g.map(plt.hist, 'var1', bins=20)