t-Distribution Principles

The Student's t-distribution emerges in statistical inference when evaluating small samples where the population standard deviation is unknown. Conceptually, if a variable $Z$ follows a standard normal distribution $N(0,1)$ and $V$ follows a chi-square distribution with $k$ degrees of freedom, independent of $Z$, then $T = Z / \sqrt{V/k}$ follows a t-distribution with $k$ degrees of freedom. As degrees of freedom increase, the t-distribution's heavier tails diminish, and its shape converges with the standard normal distribution.

The calculation for the t-statistic is defined as:

$$T = \frac{\bar{X} - \mu}{S / \sqrt{n}}$$

Where:

• $\bar{X}$: Sample mean
• $\mu$: Hypothesized population mean
• $S$: Sample standard deviation
• $n$: Number of observations

t-Test Principles

The t-test applies the t-distribution to determine if the discrepancy between sample means is statistically significant. It calculates a t-statistic that follows the t-distribution under the null hypothesis.

One-Sample t-Test Assesses if a single sample mean significantly deviates from a known or theoretical constant $\mu_0$.

$$t = \frac{\bar{x} - \mu_0}{s / \sqrt{n}}$$

Independent Two-Sample t-Test Compares the means of two distinct, unrelated groups to identify significant differences.

$$t = \frac{\bar{x}_1 - \bar{x}_2}{\sqrt{\frac{s_1^2}{n_1} + \frac{s_2^2}{n_2}}}$$

Paired Sample t-Test Evaluates related samples, such as repeated measurements on the same subjects under varying conditions.

$$t = \frac{\bar{d} - \mu_d}{s_d / \sqrt{n}}$$

Where $\bar{d}$ is the mean of the differences, $\mu_d$ is the hypothesized mean difference (often 0), and $s_d$ is the standard deviation of the differences.

R Implementation

Visualizing the t-Distribution

dof <- 8
x_sequence <- seq(-6, 6, length.out = 500)
y_density <- dt(x_sequence, dof)

plot(x_sequence, y_density, type = "l", col = "darkblue", lwd = 2,
     xlab = "t Statistic", ylab = "Density",
     main = paste("t-Distribution Shape (df =", dof, ")"))
polygon(c(min(x_sequence), x_sequence, max(x_sequence)), 
        c(0, y_density, 0), col = "lightblue", border = NA)

Plotting the Upper Tail

dof <- 8
right_x <- seq(0, 6, length.out = 500)
right_y <- dt(right_x, dof)

plot(right_x, right_y, type = "l", col = "darkblue", lwd = 2, xlim = c(-1, 6),
     xlab = "t Statistic", ylab = "Density",
     main = paste("One-Tailed Region (df =", dof, ")"))
polygon(c(0, right_x, 6), c(0, right_y, 0), col = "lightblue", border = NA)

Plotting Two-Tailed Rejection Regions

dof <- 8
full_x <- seq(-6, 6, length.out = 500)
full_y <- dt(full_x, dof)

plot(full_x, full_y, type = "l", col = "darkblue", lwd = 2,
     xlab = "t Statistic", ylab = "Density",
     main = paste("Two-Tailed Regions (df =", dof, ")"))

# Shade left tail (x < -2.5)
left_x <- full_x[full_x < -2.5]
left_y <- full_y[full_x < -2.5]
polygon(c(min(left_x), left_x, -2.5), c(0, left_y, 0), col = "blue", border = NA)

# Shade right tail (x > 2.5)
right_x <- full_x[full_x > 2.5]
right_y <- full_y[full_x > 2.5]
polygon(c(2.5, right_x, max(right_x)), c(0, right_y, 0), col = "blue", border = NA)

Executing a One-Sample t-Test

# Evaluating if observations differ from a standard value of 15
obs_values <- c(14.8, 15.2, 14.9, 15.1, 15.0, 14.7, 15.3)
result_one <- t.test(obs_values, mu = 15)
print(result_one)

The output returns a t-value of 0, demonstrating the sample mean exact matches the hypothesized mean of 15. The p-value is 1, providing no grounds to reject the null hypothesis. Degrees of freedom are calculated as $n - 1 = 6$.

Executing an Independent Two-Sample t-Test

# Comparing means between two distinct experimental groups
group_a <- c(22.1, 21.5, 22.8, 21.9, 22.3)
group_b <- c(20.5, 21.0, 20.8, 21.2, 20.6, 21.1, 20.9, 21.3)
result_two <- t.test(group_a, group_b)
print(result_two)

The Welch Two Sample t-test generates a t-value quantifying the mean difference relative to the pooled standard error. A resulting p-value below the 0.05 significance threshold indicates a statistically significant difference in group means. The degrees of freedom are adjusted via the Welch formula to account for potential variance disparities.

Executing a Paired t-Test

# Comparing pre-treatment and post-treatment measurements on the same subjects
pre_treatment <- c(8.5, 8.9, 8.2, 9.1, 8.7)
post_treatment <- c(8.0, 8.2, 7.9, 8.5, 8.1)
result_paired <- t.test(post_treatment, pre_treatment, paired = TRUE)
print(result_paired)

The paired test computes variations within each matched pair. A negative t-value signifies that the second condition's mean falls below the first. If the p-value drops beneath 0.05, the null hypothesis of zero mean difference is rejected, confirming a significant alteration between the paired conditions.