What Degrees of Freedom Tell Us
In “Why We Use t Instead of z for Means,” you saw that a one-sample t interval uses the sample standard deviation \(s\) to estimate the unknown population standard deviation \(\sigma\). The particular t distribution used depends on the sample size. Its degrees of freedom, abbreviated \(df\), determine how heavy its tails are.
For a one-sample t procedure, the degrees of freedom are \(n-1\), where \(n\) is the sample size. The subtraction of one reflects the fact that the sample mean \(\bar{x}\) is estimated from the same observations used to calculate \(s\). Once \(\bar{x}\) is known, the observations’ deviations from it must add to zero. That constraint means only \(n-1\) of those deviations can vary freely.
The idea is easiest to see with a small data set. If five observations have a known mean, their five deviations from that mean must sum to zero. If you know four deviations, the fifth is forced to be the negative of their sum. It is not free to take any value. So although there are five observations, there are only four independently varying deviations around the estimated mean.
This is an intuitive explanation of the \(n-1\) rule, not a separate method for calculating \(s\). In a one-sample t procedure, use the rule directly: subtract one from the sample size. Do not subtract one from the number of groups, or use a separate degrees-of-freedom formula from another procedure.
How Degrees of Freedom Shape the t Curve
A t distribution is centered at zero and is symmetric. Like the standard Normal distribution, it describes values on a standardized scale. For an independent random sample from a Normal population, the standardized statistic
follows a t distribution with \(n-1\) degrees of freedom. The numerator measures how far the sample mean is from the population mean; the denominator estimates the standard error of the sample mean.
The t distribution is a family of curves, not one single curve. With few degrees of freedom, a t curve has heavier tails than the standard Normal curve. In practical terms, it assigns more probability to standardized values far from zero. As \(df\) increases, the tails become less heavy and the t curve approaches the standard Normal curve. For any finite \(df\), the t curve is not exactly the standard Normal curve.
This pattern fits the reason for using t. With a small sample, \(s\) can be a less stable estimate of \(\sigma\), so there is more uncertainty in the estimated standard error. The heavier tails reflect that uncertainty. As the sample size—and therefore \(df\)—increases, the estimate of spread generally becomes more stable and the adjustment becomes smaller.
The following table gives selected two-sided 95% critical values. These are rounded values from a t distribution calculator or table. The standard Normal critical value for a central 95% area is included for comparison.
| Degrees of freedom | 95% t critical value | Comparison |
|---|---|---|
| 4 | 2.776 | Heavier tails; noticeably above 1.960 |
| 9 | 2.262 | Closer to 1.960 |
| 24 | 2.064 | Closer still to 1.960 |
| 39 | 2.023 | Only slightly above 1.960 |
| Standard Normal | 1.960 | Comparison value |
A critical value marks a cutoff on the horizontal axis that captures a specified central area under a curve. At 95% confidence, the t critical value leaves 2.5% in each tail. With fewer degrees of freedom, the cutoff must be farther from zero to capture that same central area. This is why a t interval’s margin of error is generally larger than a z-based comparison using the same estimated standard error.
Worked Examples: Finding and Using Degrees of Freedom
Worked Example: Seeing the One-Degree-of-Freedom Constraint
Five measurements of the time, in minutes, for a small device to complete a cycle are 4, 5, 6, 7, and 8. Find the sample mean, list the deviations from that mean, and explain why a one-sample t procedure has 4 degrees of freedom.
Find the mean. The sample mean is
Find the deviations. Subtract 6 from each observation. The deviations are \(-2,-1,0,1,\) and \(2\) minutes, and their sum is
Explain the degrees of freedom. Once the first four deviations are known, the fifth must make the total zero. Here, if the first four deviations sum to \(-2\), the final deviation must be \(2\). It is not independent of the others. Thus \(n=5\) observations provide \(5-1=4\) degrees of freedom for a one-sample t procedure.
Answer. The one-sample t procedure has \(df=n-1=5-1=4\). The degrees of freedom count the independent deviations around the estimated sample mean, not the total number of observations.
Worked Example: Comparing t Critical Values
Suppose we compare 95% critical values for one-sample t procedures with \(df=4\), \(df=9\), and \(df=24\). Use the critical values in the table above to describe the trend. To isolate the effect of the critical value, imagine that each comparison uses an estimated standard error of 1.2 units.
Find the critical values. For 95% confidence, the respective t critical values are 2.776, 2.262, and 2.064. The standard Normal critical value is 1.960.
Compare the margins of error. A margin of error is the critical value multiplied by the estimated standard error. For \(df=4\), it is
For \(df=9\), it is
For \(df=24\), it is
For comparison, the standard Normal critical value gives a margin of error of \(1.960(1.2)=2.352\) units.
Explain the pattern. As \(df\) rises from 4 to 24, the t critical value and the illustrative margin of error decrease toward their standard Normal comparison values. The t curve’s heavier tails are most noticeable at low degrees of freedom. The standard error of 1.2 units is held fixed here only to make the effect of the critical value easy to see; actual intervals can have different standard errors as well.
Worked Example: A One-Sample t Interval With 39 Degrees of Freedom
A technician selects a random sample of 40 devices from a very large production population. The mean cycle time is 72.4 minutes, and the sample standard deviation is 8 minutes. The sample has no strong skewness or extreme outliers. Find the degrees of freedom and use the 95% t critical value \(t^*=2.023\) to calculate an interval for the population mean cycle time.
State. The parameter is \(\mu\), the population mean cycle time for the devices. The point estimate is \(\bar{x}=72.4\) minutes. The population standard deviation is unknown, so a one-sample t interval is appropriate.
Plan. The sample is random, and 40 is less than 10% of the very large population, so the 10% condition supports treating the observations as independent. The sample has 40 observations and no strong skewness or extreme outliers, so using a t procedure is reasonable. The degrees of freedom are \(n-1=39\). Use the interval form \(\bar{x}\pm t^*(s/\sqrt{n})\).
Do. First calculate the estimated standard error:
The margin of error is
The interval is
Conclude. We are 95% confident that the population mean cycle time for these devices is between approximately 69.841 and 74.959 minutes. The 39 degrees of freedom determine the t curve used for the critical value. Because 39 is fairly large, its 95% critical value, 2.023, is relatively close to the standard Normal value, 1.960.
Common Mistakes and AP Exam Tips
- Using \(n\) instead of \(n-1\). For a one-sample t procedure, calculate \(df=n-1\). If \(n=18\), for example, \(df=17\), not 18.
- Thinking degrees of freedom are observations removed from the sample. No observation is discarded. The \(n-1\) count describes the independent information in deviations around the estimated sample mean.
- Claiming every t curve is the same. The curve changes with \(df\). State which degrees of freedom apply when identifying a t distribution or its critical value.
- Saying a t curve becomes exactly Normal for a large finite sample. It approaches the standard Normal curve as \(df\) grows. At any finite \(df\), the distributions are not exactly identical.
- Assuming a larger sample changes the definition of the parameter or the procedure. The target remains the population mean \(\mu\), and when \(\sigma\) is unknown the one-sample t procedure still uses \(s\). Increasing \(n\) changes \(df\) and generally brings the t critical value closer to the standard Normal critical value.
- Giving only a number without connecting it to the setting. A clear response says, for example, “With \(n=40\), the one-sample t procedure has \(df=40-1=39\).” In an interval problem, also identify the population mean and report the interval in the context and units.
Keep the curve’s shape separate from the shape of the population data. A t distribution is the model for a standardized statistic under the assumptions of the procedure; it does not mean the individual measurements themselves follow a t distribution. The data conditions still matter, especially for a small sample.
Check Your Understanding
Use the one-sample rule \(df=n-1\) and explain what the degrees of freedom imply about the t curve.
- A random sample of 12 batteries is used in a one-sample t procedure. What are the degrees of freedom?
- In a sample of six measurements with a known sample mean, why is the sixth deviation from that mean determined once the other five deviations are known?
- Which has the heavier tails: a t distribution with \(df=5\) or one with \(df=30\)? Explain.
- As degrees of freedom increase, what happens to the t curve and its critical values compared with the standard Normal curve?
- A student says that a one-sample t procedure with \(n=25\) has 25 degrees of freedom because there are 25 observations. Correct the statement and explain the rule.