91Ó°ÊÓ

In statistics, the t-test is a powerful tool used to determine if there is a significant difference between the means of two groups. Perfectly suited for small sample sizes, it helps us decide whether any observed difference is meaningful or likely occurred by chance.
The t-test comes in various forms, but when comparing means, it usually involves:

• A sample mean (\( \bar{x} \)): your calculated average from sample data.
• A known mean or hypothesized value (\( \mu \)): the mean you're testing against, often from the null hypothesis.
• The sample standard deviation (\( s \)): which gives insight into data variability.
• The sample size (\( n \)): the number of observations in your sample.

The result of a t-test is the t-statistic, which tells us how many standard deviations the sample mean is from the known mean. If this value is far from zero, it suggests a difference worth noting.
It’s essential to also consider the critical value, which depends on your confidence level and degrees of freedom. If the t-statistic exceeds this, our observations are statistically significant. Otherwise, they might not substantially differ.

The null hypothesis acts as a starting assumption in hypothesis testing, often suggesting no effect or no difference exists. In our exercise, the null hypothesis (\( H_0 \)) posits that the mean difference between the two voltmeters' measurements is zero, meaning there is no calibration discrepancy.
This hypothesis is vital because it forms the basis of statistical testing:

• If evidence against \( H_0 \) is strong enough, we reject it.
• If not, we fail to reject it, not necessarily proving it true, but accepting it based on insufficient contrary evidence.

Rejecting the null hypothesis implies the alternate hypothesis (\( H_a \)) might be true: in our exercise, that a calibration difference indeed exists. Consequently, deciding whether to reject \( H_0 \) involves calculating a test statistic and comparing it against a critical threshold. If our statistic exceeds this threshold, the observations provide enough evidence to suggest a meaningful deviation from the null assumption.

The sample mean is a core statistical measure representing the average within a dataset, calculated by adding all observations and dividing by the total count. In hypothesis testing, the sample mean (\( \bar{x} \)) offers a benchmark to compare with an expected value under the null hypothesis.
In our exercise about voltmeters, the differences in measurements were summed and divided by the number of observations to find the mean:

• Sum of differences: \( 0.8 + 0.2 - 0.3 + 0.1 + 0.0 + 0.0 + 0.5 + 0.7 + 0.2 = 2.2 \)
• Number of observations: 9

Thus, the sample mean becomes \( \bar{x} = \frac{2.2}{9} \approx 0.244 \).
This average difference forms the basis for further comparison with the null hypothesis, helping to determine if the observed discrepancy is statistically noteworthy or just a result of randomness.

Standard deviation is a measure of data variability or dispersion around the mean. It tells us how much the individual data points differ from the average value. Knowing the standard deviation is key to understanding the test statistic in our t-test.
In our example with voltmeter readings, we first calculate the squared differences between each value and the sample mean, sum them up, and divide by one less than the number of observations (degree of freedom), before taking the square root:

• Calculate squared differences: e.g., \( (0.8 - 0.244)^2, (0.2 - 0.244)^2, ext{etc.} \)
• Sum these values and divide by \( n-1 \) (here, 8), then compute the square root.

The standard deviation helps measure how spread out the differences are, key to computing the t-statistic as it influences how sharply or widely scattered our data is perceived, thus affecting the outcome of our hypothesis test.