t-distribution confidence interval: Used when population standard deviation σ is unknown and sample size is small (n < 30). Formula: X̄ ± t*α/2, df(s/√n), where df = n-1.
- Identify the sample statistics: X̄, s, n
- Determine the confidence level and degrees of freedom: df = n-1
- Find the critical t-value: tα/2, df
- Calculate the standard error: s/√n
- Calculate the margin of error: tα/2, df(s/√n)
- Construct the confidence interval: X̄ ± margin of error
X̄ = 78 (sample mean), s = 12 (sample standard deviation), n = 15 (sample size)
df = n - 1 = 15 - 1 = 14
For 95% confidence, α = 0.05, α/2 = 0.025, t0.025, 14 = 2.145
SE = s/√n = 12/√15 = 12/3.873 ≈ 3.098
ME = t × SE = 2.145 × 3.098 ≈ 6.646
CI = X̄ ± ME = 78 ± 6.646 = [71.354, 84.646]
We are 95% confident that the population mean test score is between 71.35 and 84.65.
• t-interval formula: X̄ ± tα/2, df(s/√n) when σ is unknown
• Degrees of freedom: df = n - 1
• Small sample correction: t-distribution has heavier tails than normal distribution
• Use t-distribution when σ is unknown and n < 30
• The t-distribution has more spread than normal distribution for small samples
• As df increases, t-distribution approaches normal distribution
Confidence level effect on t-intervals: Higher confidence levels result in wider intervals because they require larger critical t-values to capture more of the sampling distribution.
SE = s/√n = 2.3/√20 = 2.3/4.472 ≈ 0.514
df = n - 1 = 20 - 1 = 19
90% confidence: t0.05, 19 = 1.729
95% confidence: t0.025, 19 = 2.093
99% confidence: t0.005, 19 = 2.861
90%: ME = 1.729 × 0.514 ≈ 0.889
95%: ME = 2.093 × 0.514 ≈ 1.076
99%: ME = 2.861 × 0.514 ≈ 1.471
90% CI: 8.5 ± 0.889 = [7.611, 9.389]
95% CI: 8.5 ± 1.076 = [7.424, 9.576]
99% CI: 8.5 ± 1.471 = [7.029, 9.971]
Higher confidence levels produce wider intervals. The 99% CI is widest, followed by 95%, then 90%. This reflects the trade-off between confidence and precision.
• Confidence-precision tradeoff: Higher confidence requires wider intervals
• t-values: Increase as confidence level increases
• Interval width: Proportional to the critical t-value
• Higher confidence levels increase certainty but decrease precision
• The interval width increases as the critical t-value increases
• Medical research often uses 95% confidence as a standard
Degrees of freedom effect: The t-distribution approaches the standard normal distribution as degrees of freedom increase. Critical t-values decrease toward z-values as sample size increases.
n = 5: df = 4
n = 15: df = 14
n = 30: df = 29
t0.025, 4 = 2.776
t0.025, 14 = 2.145
t0.025, 29 = 2.045
z0.025 = 1.96
As df increases, t-values approach the z-value of 1.96
As sample size increases, critical t-values decrease and approach the z-value of 1.96. This shows that the t-distribution converges to the normal distribution as sample size increases.
• Convergence property: t-distribution approaches normal distribution as df increases
• Small sample correction: t-values are larger than z-values for small samples
• Degrees of freedom: df = n - 1 determines the shape of the t-distribution
• For large samples (n > 30), t-values are very close to z-values
• The t-distribution accounts for the extra uncertainty when estimating σ
• This convergence justifies using normal distribution for large samples
t-distribution: A family of distributions used when the population standard deviation is unknown and sample size is small
Degrees of Freedom (df): The parameter that determines the shape of the t-distribution, typically df = n - 1
Small Sample: Typically defined as n < 30, requiring the use of t-distribution instead of normal distribution
Sample Standard Deviation (s): Used to estimate the population standard deviation σ when it's unknown
- Check conditions: σ unknown, normal population or large sample
- Identify parameters: X̄, s, n
- Calculate degrees of freedom: df = n - 1
- Find critical t-value: Look up tα/2, df for desired confidence level
- Calculate standard error: SE = s/√n
- Calculate margin of error: ME = tα/2, df × SE
- Construct interval: CI = X̄ ± ME
• Interval width: 2 × ME = 2 × tα/2, df(s/√n)
• Degrees of freedom: df = n - 1
• As df → ∞, t-distribution → normal distribution
• Confidence-precision tradeoff: Higher confidence → Wider interval
Sample size effect on t-intervals: Larger samples produce narrower confidence intervals due to both reduced standard error and smaller critical t-values.
df₁ = 7, t0.025, 7 = 2.365
SE₁ = 40/√8 = 40/2.828 ≈ 14.14
ME₁ = 2.365 × 14.14 ≈ 33.4
CI₁ = 250 ± 33.4 = [216.6, 283.4]
df₂ = 24, t0.025, 24 = 2.064
SE₂ = 40/√25 = 40/5 = 8
ME₂ = 2.064 × 8 ≈ 16.5
CI₂ = 250 ± 16.5 = [233.5, 266.5]
Width₁ = 283.4 - 216.6 = 66.8 ms
Width₂ = 266.5 - 233.5 = 33.0 ms
The interval for n = 8 is wider (66.8 ms) than for n = 25 (33.0 ms). Larger samples provide more precise estimates with narrower confidence intervals.
• Sample size effect: Larger samples → smaller standard errors → narrower intervals
• Degrees of freedom: More df → smaller critical t-values → narrower intervals
• Quadratic relationship: To halve the width, quadruple the sample size
• The width decreased by more than half when sample size increased from 8 to 25
• Both the reduced standard error and the smaller t-value contributed to the narrower interval
• This demonstrates why researchers prefer larger samples when possible
Practical interpretation: Using confidence intervals to assess claims about population parameters in real-world contexts.
df = n - 1 = 12 - 1 = 11
t0.005, 11 = 3.106
SE = s/√n = 8/√12 = 8/3.464 ≈ 2.31
ME = t × SE = 3.106 × 2.31 ≈ 7.17
CI = X̄ ± ME = 45 ± 7.17 = [37.83, 52.17]
Since 50 is within [37.83, 52.17], it is a plausible value for the true mean.
The 99% confidence interval for the true mean battery life is [37.83, 52.17] hours. Since 50 hours falls within this interval, it is a plausible value for the true mean battery life.
• Plausibility assessment: Values within the confidence interval are considered plausible
• High confidence: 99% confidence provides strong evidence for the range of plausible values
• Quality control: Confidence intervals help assess product performance specifications
• The wide interval reflects both the small sample size and high confidence level
• The company would need a larger sample to get a more precise estimate
• This type of analysis is fundamental in product development and quality assurance
Comprehensive Summary: Confidence Interval for Population Mean (t-distribution intro)
t-distribution: A family of continuous probability distributions that arise when estimating the mean of a normally distributed population when the sample size is small and the population standard deviation is unknown.
Degrees of Freedom (df): The parameter that determines the shape of the t-distribution, typically df = n - 1 for a single sample.
Confidence Interval using t-distribution: An interval estimate of the population mean when σ is unknown: CI = X̄ ± tα/2, df(s/√n).
Small Sample: Conventionally defined as n < 30, requiring the use of t-distribution instead of normal distribution for confidence intervals.
- When to use t-distribution: When σ is unknown and n < 30 (or population is normal with unknown σ)
- Basic Formula: CI = X̄ ± tα/2, df(s/√n)
- Degrees of Freedom: df = n - 1
- Comparison to z-distribution: t-distribution has heavier tails, providing wider intervals for small samples
- Convergence: As df increases, t-distribution approaches normal distribution
Sample: n = 10, X̄ = 50, s = 8
df = 9, t0.025, 9 = 2.262
SE = 8/√10 = 2.53
95% CI: 50 ± 2.262 × 2.53 = 50 ± 5.72 = [44.28, 55.72]
Same X̄ = 50, s = 8, but different n values:
n = 5: df = 4, t0.025, 4 = 2.776, SE = 3.58, CI = [42.06, 57.94]
n = 20: df = 19, t0.025, 19 = 2.093, SE = 1.79, CI = [46.26, 53.74]
Notice how the interval narrows as sample size increases.
• Confidence Interval: CI = X̄ ± tα/2, df(s/√n)
• Standard Error: SE = s/√n
• Degrees of Freedom: df = n - 1
• Margin of Error: ME = tα/2, df(s/√n)
- t-Interval Structure: Center ± Spread → X̄ ± tα/2, df(s/√n)
- When to Use: σ Unknown + Small Sample (n < 30)
- Degrees of Freedom: df = n - 1
- Heavy Tails: t-distribution has more spread than normal for small samples
Questions & Answers
Question: Why do we need the t-distribution? Why can't we just use the normal distribution all the time?
Answer: We need the t-distribution because it accounts for the additional uncertainty introduced when we estimate the population standard deviation σ using the sample standard deviation s.
When σ is known, we can use the normal distribution directly. However, when σ is unknown (which is often the case in practice), we substitute s for σ, which introduces extra variability into our estimate.
The t-distribution has heavier tails than the normal distribution, which means it's more spread out. This extra spread accounts for the uncertainty in estimating σ from the sample. As our sample size increases, our estimate of σ becomes more reliable, and the t-distribution approaches the normal distribution.
For example, with a sample size of 5, the critical t-value for 95% confidence is 2.776, while the z-value is only 1.96. The wider t-interval reflects the greater uncertainty with small samples.
Question: When exactly should I use the t-distribution versus the normal distribution for confidence intervals?
Answer: Here's the decision rule:
Use t-distribution when:
- Population standard deviation σ is unknown (most common scenario)
- Sample size is small (n < 30) AND population is normally distributed
- Or sample size is large (n ≥ 30) but σ is still unknown
Use normal distribution when:
- Population standard deviation σ is known (rare in practice)
- Sample size is large (n ≥ 30) and using sample standard deviation s
In practice, the t-distribution is used much more frequently because σ is rarely known. Even for large samples, the t-distribution is technically more appropriate when using s to estimate σ, though the difference becomes negligible.
Question: What are degrees of freedom and why do we use n-1?
Answer: Degrees of freedom (df) represent the number of independent pieces of information available to estimate a parameter.
For a sample of size n, we use df = n - 1 when calculating the sample standard deviation because:
- When we calculate s, we first calculate the sample mean X̄
- This uses up one piece of information from our sample
- Only n - 1 values can vary freely once X̄ is known
For example, if we have 5 numbers with a known mean of 10, only 4 of them can be freely chosen. The fifth is determined by the constraint that the mean must be 10.
The degrees of freedom determine the shape of the t-distribution. As df increases, the t-distribution approaches the normal distribution. This is why t-distributions with higher df have thinner tails.
Question: How does the t-distribution differ from the normal distribution graphically?
Answer: The t-distribution differs from the normal distribution in several ways:
Shape: The t-distribution is bell-shaped and symmetric like the normal distribution, but:
- It has heavier tails (more probability in the tails)
- It's more spread out (higher variance)
- It has a lower peak than the normal distribution
Dependence on df: The shape of the t-distribution depends on degrees of freedom:
- Low df: Much more spread out than normal distribution
- High df: Very close to normal distribution
The heavier tails of the t-distribution account for the extra uncertainty when estimating σ from the sample. As df increases, the t-distribution converges to the normal distribution.
For df > 30, the t-distribution is very close to the normal distribution.
Question: How do I interpret a confidence interval constructed using the t-distribution?
Answer: The interpretation of a t-based confidence interval is the same as for any confidence interval:
If you have a 95% confidence interval for battery life of [37.8, 52.2] hours, you can say:
"We are 95% confident that the true mean battery life is between 37.8 and 52.2 hours."
Practical interpretations might include:
- Whether the interval includes meaningful benchmarks (e.g., advertised life of 50 hours)
- How wide the interval is (narrow = precise estimate)
- Whether the entire interval is above or below important thresholds
The t-distribution just means we're accounting for the extra uncertainty from not knowing the population standard deviation, resulting in slightly wider intervals than if we knew σ.
The wider interval reflects the fact that we're working with less information about the population variability.