Basics of ML & AL Week 2 Challenge

2)The expected value and the simple mean (arithmetic mean) both provide measures of central tendency, but they are used in different contexts and have distinct interpretations. Let's explore their significance and how they differ:

1. Simple Mean (Arithmetic Mean): The arithmetic mean is the sum of all observed values divided by the number of observations.

2. Expected Value (Mean of a Probability Distribution): The expected value is a weighted average of all possible outcomes in a probability distribution, where each outcome is weighted by its probability of occurring.

3)Having skewness in a data distribution is not inherently "bad" or "good"; it depends on the context of the analysis and the goals of the investigation. Skewness simply provides information about the asymmetry of the distribution, and different types of skewness can have different implications depending on the field you're working in.

Types of Skewness:

• Positive skew (Right-skewed): The right tail is longer; most values are concentrated on the left, but there are some large values pulling the mean to the right.
• Negative skew (Left-skewed): The left tail is longer; most values are concentrated on the right, but there are some small values pulling the mean to the left.

Significance of Skewness in Analysis:

1. In Statistical Assumptions:

   • Normality assumption: Many statistical tests (e.g., t-tests, ANOVA, linear regression) assume that the data is normally distributed (no skewness, or skewness ≈ 0). If the data is heavily skewed, these tests may yield inaccurate results.
   • Impact: A high level of skewness may violate the normality assumption, leading to:
     • Biased estimates of parameters.
     • Incorrect p-values or confidence intervals.
     • Poor model fit in predictive models.

   What to do: If skewness is an issue for the method you're using, you can:

   • Transform the data (e.g., log transformation, square root transformation) to reduce skewness.
   • Use non-parametric tests that don’t rely on normality assumptions (e.g., Mann-Whitney U test, Kruskal-Wallis test).

2. In Descriptive Analysis:

   • Interpreting the data: Skewness can help you better understand the nature of your data. For example:
     • Positive skew: May indicate a few extreme values (outliers) that are pulling the average up (e.g., very high incomes in a wealth distribution).
     • Negative skew: Could indicate a concentration of high values with a few lower outliers (e.g., exam scores where most students scored high, but a few scored low).

   In these cases, skewness can provide insights about outliers or whether the data is concentrated in specific ranges.

3. In Business/Financial Analysis:

   • Positive skew: Often seen in sales, income, and asset returns, where the majority of the values are low, but a few outliers (e.g., extremely high sales or returns) skew the distribution to the right.
     • Implication: Positive skewness can indicate potential high rewards (e.g., an investment with a small chance of a big payoff).
   • Negative skew: In finance, negative skewness can indicate the risk of extreme negative returns, which could be undesirable depending on the risk tolerance.
     • Implication: Negative skewness may highlight the risk of large losses (e.g., a financial portfolio may have a large probability of small gains but also a small probability of very large losses).

4. Impact on Mean vs. Median:

   • In a skewed distribution, the mean is pulled towards the longer tail (extreme values), while the median (middle value) remains closer to where the bulk of the data is.
     • Positively skewed data: Mean > Median
     • Negatively skewed data: Mean < Median
   • Implication: In some cases (e.g., salary distributions, real estate prices), the median is a better measure of central tendency than the mean because it is less affected by extreme outliers.

When Is Skewness Considered "Bad"?

1. When Normality Is Crucial:

   • In methods that require normal distribution, such as certain statistical models, high skewness may reduce the validity of the results.
   • Example: If you’re building a linear regression model, strong skewness in residuals might indicate that the model is not well-fitted, and the assumptions are violated.

2. When Predictive Accuracy Suffers:

   • If skewness affects the ability to predict or generalize findings from a model (e.g., in machine learning), it might be a problem. For instance, skewness in the target variable can make models like linear regression perform poorly.

3. In Finance (especially negative skew):

• Investors often dislike negative skewness because it suggests a higher likelihood of extreme losses. For instance, a negatively skewed investment return means that while small positive returns are more frequent, there’s a higher risk of rare, significant losses.