Skip to main content

Relevant Statistics

Updated Mar 03, 2023 ·

Overview​

When comparing data, we can choose between different ways to express changes or patterns. Depending on the question, we can use either absolute numbers or growth percentages to represent changes.

  • Absolute Change

    • Used when comparing a single variable.
    • Shows the total difference (e.g., 3,000 more units sold).

  • Relative Change

    • Useful when comparing several variables.
    • Shows percentage growth (e.g., 10% more units sold).
    • Helps compare variables of different scales (e.g., sales in thousands vs. millions).

Proportional Changes​

Sometimes, we want to look at how the proportion of sales changed between years.

Proportional Comparison

  • Comparing the percentage change in sales helps highlight shifts.
  • Example: When one product grows while another decreases.
  • Shows meaningful changes in volume and helps compare product sales.

Ratios​

A ratio is useful for comparing two variables, particularly when the data has different scales.

  • A ratio compares two numbers as a quotient (e.g., revenue per customer).
  • Helps normalize data for better comparisons
  • Example: Comparing sales across different products.

Aggregates​

Aggregates summarize data into a single representative value.

  • Total or Count

    • Simple metrics like total sales or campaign duration.
    • Provides an overview.

  • Mean

    • The average number can represent trends.
    • Easy to calculate but can be misleading if there are outliers.
    • Example: Average chocolates or chips sold per year.

  • Median

    • The middle value in a dataset.
    • More reliable when there are extreme values or outliers.
    • Example: In 2019, the US average wage was $51,916, but the median was $34,248, showing the real distribution.

P-Value​

In statistical analysis, we often use p-values to determine the significance of our results.

  • A p-value below 0.05 suggests the result is statistically significant.
  • Lower p-value indicates stronger evidence, but it doesn’t prove the truth — just that the result is unlikely due to randomness.

To avoid confusion, consider providing additional metrics or alternative tests.