Statistical Analysis and Interpretation Notes

Key Points on Statistical Analysis and Interpretation

Basic Statistical Concepts

  • P-Value

    • A p-value indicates the significance of results in hypothesis testing.

    • If p-value < 0.05: Statistically significant (not zero); we can reject the null hypothesis.

    • If p-value > 0.05: Not statistically significant (could be zero); we fail to reject the null hypothesis.

    • Example:

    • P-value of 0.38 indicates uncertainty about an estimate of 0.44 (it may vary and could include zero).

    • An estimate of 2.8 suggests confidence that it's statistically nonzero.

Understanding Regression Outputs

  • Linear Regression Results

    • Mention the importance of clearly stating variables in your comparisons (for example, in trampoline stiffness).

    • Should specify the only difference between the groups being compared (e.g., spring stiffness in trampolines).

    • Statement format:

    • "For every one unit increase, the response (e.g., bounce height) increases by X on average."

    • Emphasize that context matters—ensure clarity for assumptions made (e.g., comparing trampoline stiffness alone).

Coefficient of Determination (R-Squared)

  • R-Squared (R²)

    • R² represents the proportion of variance explained by the model.

    • Found by squaring the correlation coefficient (r): R2=r2R^2 = r^2

    • If R² = 0.5,

    • 50% explained variance—considered acceptable.

    • Low R² values (e.g., 20%) suggest a poor model.

Making Predictions

  • Extrapolation vs. Interpolation

    • Predicting beyond existing data (extrapolation) can lead to unreliable results.

    • Example given: A minor change in stiffness leading to a significant jump height may seem unrealistic when predicted from a model.

    • Caution: Very strong outputs from models can indicate extrapolation outside reliable data range, hence predictions should be approached with skepticism.

Graph Interpretation

  • Annotating Graphs

    • Essential for clear interpretation: label axes, identify study participants, and the context of the graph.

  • Understanding Variables

    • Clarify demographics vs. performance position variables:

    • Consider the relationships shown in graphs; identify the independent and dependent variables accurately.

    • Recognize categories within variables (e.g., white women, women of color) when analyzing distribution.

  • Types of Distributions

    • Conditional Distribution:

    • Probability of a demographic within a specific context (e.g., entry-level positions among women of color).

    • Not a histogram as there are gaps in data; distributions should be continuous without gaps for such classification.

Validity of Assumptions and Statistical Conclusions

  • Summing Probabilities

    • Conditional distributions must sum to 100%, and graphs must be interpreted with caution concerning the number of people in different employment hierarchies (entry-level vs. executive roles).

    • Avoid incorrect arithmetic sum of different categories—conditions might vary significantly from level to level and should be assessed with contextual understanding rather than numerical addition alone.

Final Checklist for Reports

  • Annotate graphs with contextual backgrounds.

  • Ensure a thorough analysis of variable differences.

  • Clearly articulate assumptions and limits of your model and predictions.

  • Pay attention to detailed formatting when stating findings to maintain clarity and enhance understanding.