Detailed Study Notes on F-Test and T-Test in Statistical Analysis

Introduction to Statistical Testing

• Overview of statistical tests: F-test and T-test.

• Purpose: Evaluating differences between datasets.

F-Test

• Initial step before conducting a T-test.

• Compares the variances (or standard deviations) of two datasets.

Calculation of F-Test

• Denote standard deviations:

  • Let $ heta1$ be the standard deviation of dataset one ($s</em>1$).

  • Let $ heta2$ be the standard deviation of dataset two ($s</em>2$).

• If datasets have very different standard deviations (narrow vs. wide measurements):

  • Approach: Calculate the F-ratio.

  • Formula: $F<em>{calc} = \frac{s</em>1^2}{s_2^2}$

  • Place the larger standard deviation on top.

F-Test Outcomes

• Result Comparison: Compare F-calculated value ($F<em>{calc}$) against F-table value ($F</em>{table}$).

  • If $F<em>{calc} > F</em>{table}$: F-test fails.

  • If $F<em>{calc} < F</em>{table}$: F-test passes.

• Implication of results:

  • High F-calculated value indicates significant difference between standard deviations.

Significance of F-Test

• A F-test is necessary to determine whether to use pooled variances for T-test calculations.

Transition to T-Test

• If F-test is passed, T-test calculations proceed:

  • Denote T-calculated value ($T<em>{calc}$) and T-table value ($T</em>{table}$).

  • If $T<em>{calc} < T</em>{table}$: Datasets agree.

  • If $T<em>{calc} > T</em>{table}$: Datasets do not agree.

Example: Arsenic Measurements

• Measurement series from two different methods.

• Dataset 1: Mean = 11.87, Standard deviation = 0.31.

• Dataset 2: (assuming more spread in data) using a different measurement technique.

Detailed Calculation Steps

F-Test Calculation

• Determine Standard Deviations:

  • Dataset 1: $s<em>1 = 0.80$, Dataset 2: $s</em>2 = 0.31$.

• Check Degrees of Freedom:

  • Each dataset has three measurements; thus degrees of freedom is:

  • For each dataset: $n-1 = 2$.

  • Total: $2 + 2 = 4$.

• Use $s<em>1$ and $s</em>2$ consistency check:

  • $F_{calc} = \frac{(0.80)^2}{(0.31)^2} \approx 7.1$

• F-Table Value:

  • Using degrees of freedom (2 for both datasets): Critical F-value reported as 19.

Interpretation of F-Test

• Compare values:

  • $7.1 < 19$: Passes F-test.

  • Conclusion: The difference in standard deviations is not significant.

Transitioning to T-Test

• Calculate pooled standard deviation ($S_{pool}$):

  • Formula for $S<em>{pool}$: $S</em>{pool} = \sqrt{\frac{(s<em>1^2 \cdot (n</em>1 - 1) + s<em>2^2 \cdot (n</em>2 - 1))}{n<em>1 + n</em>2 - 2}}$

  • Where $n<em>1$ and $n</em>2$ are the number of measurements for each dataset.

• Average two standard deviations:

  • Result should lie between both standard deviations calculated earlier.

Calculating T-Test

• Find $T_{calc}$ using:

  • $T<em>{calc} = \frac{(x</em>{bar1} - x<em>{bar2})}{S</em>{pool} \cdot \sqrt{\frac{1}{n<em>1} + \frac{1}{n</em>2}}}$

  • Resulting value $<br>ightarrow 0.285$.

• Compare against $T_{table}$ (using 4 degrees of freedom): Critical value found as 2.776.

• Final Interpretation:

  • Since $0.285 < 2.776$: Datasets agree.

Confidence Intervals

• Highlight last step: Calculate 95% confidence intervals for both datasets to check for overlap.

Conclusion

• Importance of understanding the F-test and T-test interconnectedness.

• Practical implications for evaluating measurement methods in research.