Critical Value and Confidence Intervals
Critical Value
Excel outputs critical values during hypothesis testing, labeled as "crit" or "Critical".
Critical value: Dividing point between the rejection and non-rejection regions.
Depends on:
Distribution (t, Z, etc.)
Degrees of freedom (for t)
Significance level \,\alpha
Direction of alternative hypothesis
Critical Value Notation
Z{\alpha/2} and t{\alpha/2} for two-tailed tests.
Z{\alpha} and t{\alpha} for one-tailed tests.
Excel provides positive critical values; user accounts for direction.
Obtaining Critical Values
Online calculators.
Excel:
norm.inv()for standard normal Z.t.inv()for left/right-tailed t-tests.t.inv.2t()for two-tailed t-tests.
Tables (e.g., Appendix B).
Z Critical Values in Excel
Use
norm.inv().Left-tailed:
norm.inv(\alpha, 0, 1).Right-tailed:
norm.inv(1-\alpha, 0, 1).Two-tailed:
norm.inv(\alpha/2, 0, 1)(lower); upper is the positive version.
Finding t Critical Values
Historically via tables (Appendix B).
Excel:
t.inv()andt.inv.2t()functions.
Excel Two-Tailed t Critical Values
Use
t.inv.2t(\alpha, DF).\,DF = n - 1
Excel One-Tailed t Critical Values
Use
t.inv().Left-tailed:
t.inv(\alpha, DF).Right-tailed:
t.inv(1-\alpha, DF).
Critical Value Decision Making
Is the test statistic more extreme than the critical value?
Yes: Reject the null hypothesis.
No: Do not reject the null hypothesis.
Consistent with P-value approach.
Estimation and Confidence Intervals
Point Estimate: Single approximate value for a population parameter.
Interval Estimate: Range of values likely to contain the population parameter.
Confidence Interval (CI): Range of numbers from sample data likely to include the true population parameter; an empirical interval estimate.
Two Sided Intervals
Most CIs are two-sided: point estimate ± margin of error.
Margin of error: Half-width of a CI; measure of uncertainty.
Margin of error = critical value * standard error.
Math for the CI
Statistic from data is at the center of the sampling distribution.
Use the sampling distribution to create an interval likely to contain the true population parameter.
Using the CLT
CLT: Central Limit Theorem.
CLT Condition: The population must be normally distributed or the sample size must be at least 30.
Starting from the CLT
Assume population standard deviation \sigma known.
Formula: \,\mu = \bar{x} \pm Z * \frac{\sigma}{\sqrt{n}}
Selecting Z
Pick low and high side Z values to capture a large percentage (confidence level) of the sampling distribution.
Common confidence levels: 90%, 95%, 99% (align with \,\alpha levels 0.10, 0.05, 0.01).
Building a Confidence Interval
For a symmetric 95% CI, use Z = ±1.96.
Interval: \,\bar{x} \pm 1.96 * \frac{\sigma}{\sqrt{n}}
Working through Fall 2020 Class Heights
Example: N = 53, sample mean = 67.75 inches, assume \,\sigma = 4.5 inches.
95% CI: 67.75 ± 1.96 * 4.5/\,\sqrt{53} = 66.539 to 68.961 inches.
Lower and Upper Bounds
\,\bar{x} - Z{\alpha/2} * \frac{\sigma}{\sqrt{n}} < \mu < \bar{x} + Z{\alpha/2} * \frac{\sigma}{\sqrt{n}}
Standard Error and Critical Value
Margin of error Z_{\alpha/2} * \frac{\sigma}{\sqrt{n}}. \,\frac{\sigma}{\sqrt{n}}
is standard error.
Z_{\alpha/2} is a critical value.
About Levels and alpha
CI abbreviated, like 95% CI.
Higher confidence means larger Z and wider CI.
Unknown Population Standard Deviation
Substitute sample standard deviation s for \sigma.
Use Student’s t distribution with (n – 1) degrees of freedom.
\bar{x} \pm t_{\alpha/2} * \frac{s}{\sqrt{n}}
Recap of Computational Terms
Confidence Interval = Point Estimate ± Margin of Error.
Margin of Error = Critical Value * Standard Error
Confidence Level = 1 – \,\alpha
\,\bar{x} \pm Z_{\alpha/2} * \frac{\sigma}{\sqrt{n}},
if population \sigma known
\,\bar{x} \pm t_{\alpha/2} * \frac{s}{\sqrt{n}}, if population \sigma not known with (n – 1) DF
Converting Between The Forms
Lower Bound = Point Estimate – Margin of Error
Upper Bound = Point Estimate + Margin of Error
Point Estimate = (Lower Bound + Upper Bound) / 2
Margin of Error = Upper Bound – Point Estimate
Assigning Probability
CI for the mean is sample mean ± margin of error.
Normal Approximation Interval
Conditions for CI for Proportion
Sample size n is fixed and known.
There are exactly two options for each participant, occurrence or non-occurrence.
Each participant’s response is independent of all other participant’s responses.
The sample contains at least 5 occurrences and 5 non-occurrences.
Using Sample Proportion
Sample proportion \,\hat{p} = X/n.
Standard error \,\sqrt{\frac{\hat{p}*(1-\hat{p})}{n}}.
Critical value uses standard normal Z.
\,\hat{p} \pm Z_{\alpha/2} * \sqrt{\frac{\hat{p}*(1-\hat{p})}{n}}
Support or Reject?
Reject claimed values outside confidence interval.
Support claimed values inside confidence interval.
Claims are Hypotheses
H0: p = value using p for population proportion
HA: p ≠ value
Reject H0 if hypothesis value not inside confidence interval. Do not reject (support) H0 if hypothesis value is inside confidence interval.