Medical Research
- Clinical trial efficacy
- Treatment vs control
- Drug effectiveness
Determine if treatments work better than placebo
Calculate p-value from Z-score, t-statistic, F-statistic, chi-square statistic, or correlation coefficient.
https://www.solvezi.com/statistics/p-value-calculator
Determine statistical significance of test results
Common values: 1.645 (90%), 1.96 (95%), 2.576 (99%)
Interpretation:
✗ Fail to reject null hypothesis (H₀) - Results are not statistically significant
Significance Levels
Significant (*)
Very significant (**)
Extremely significant (***)
Not significant
Common Critical Values (α = 0.05, two-tailed)
z = 1.96
t = 2.228
t = 2.042
χ² = 3.841
What is a P-Value?
• P-value is the probability of obtaining test results at least as extreme as observed, assuming H₀ is true.
• Small p-value (p < 0.05): Strong evidence against H₀ → Reject H₀
• Large p-value (p ≥ 0.05): Weak evidence against H₀ → Fail to reject H₀
• Interpretation: Lower p-values indicate stronger statistical significance.
P-Value Calculator • Determine statistical significance of test results
Master p-values and statistical significance — understand hypothesis testing, interpret Z-tests, T-tests, and Chi-square tests, avoid common pitfalls, and make data-driven decisions in research, A/B testing, and quality control.
A p-value is the probability of obtaining test results at least as extreme as the observed results, assuming the null hypothesis (H₀) is true.
"Small p-value (p < 0.05) = Strong evidence against H₀ → Reject H₀
Large p-value (p ≥ 0.05) = Weak evidence against H₀ → Fail to reject H₀"
Very strong evidence against H₀
Strong evidence against H₀
Moderate evidence against H₀
Weak evidence, interpret cautiously
Insufficient evidence against H₀
z = (x̄ - μ) / (σ/√n)When to use: Large sample (n > 30), known population variance
Assumption: Data approximately normal, variance known
Common use: Comparing sample mean to population mean
t = (x̄ - μ) / (s/√n)When to use: Small sample (n ≤ 30), unknown population variance
Assumption: Data approximately normal, variance unknown
Common use: Comparing means when σ is unknown
χ² = Σ (O - E)² / EWhen to use: Categorical data, counts/frequencies
Assumption: Expected frequencies ≥ 5
Common use: Testing independence or goodness-of-fit
H₁: μ ≠ μ₀Testing for any difference (increase or decrease)
p = 2 × P(Z > |z|)
z = ±1.96 (α=0.05)
H₁: μ < μ₀Testing for decrease only
p = P(Z < z)
z = -1.645 (α=0.05)
H₁: μ > μ₀Testing for increase only
p = P(Z > z)
z = +1.645 (α=0.05)
Important: One-tailed tests should only be used when you have a strong directional hypothesis before collecting data. Two-tailed tests are more conservative and commonly used.
| Test | Alpha (α) | Two-Tailed | One-Tailed | Note |
|---|---|---|---|---|
| Z-Test | 0.05 | ±1.960 | ±1.645 | Large samples |
| Z-Test | 0.01 | ±2.576 | ±2.326 | Large samples |
| T-Test | 0.05 | ±2.228 | ±1.812 | df = 10 |
| T-Test | 0.05 | ±2.042 | ±1.699 | df = 30 |
| Chi-Square | 0.05 | 3.841 | N/A | df = 1 |
| Chi-Square | 0.05 | 5.991 | N/A | df = 2 |
T-test/Mean differences
Correlation
Chi-square
ANOVA
Remember:Statistical significance (p < 0.05) doesn't guarantee practical importance. Always report effect sizes alongside p-values!
H₀ (null): No effect/difference. H₁ (alternative): Effect exists
Typically 0.05 (5% risk of Type I error)
Based on data type and assumptions (Z/T/Chi-square)
Compute z, t, or χ² from sample data
Probability of observed/extreme results if H₀ is true
Reject H₀ if p < α; fail to reject if p ≥ α
Report p-value, effect size, and practical implications
Determine if treatments work better than placebo
Compare performance between variants
Verify quality improvements
Test hypotheses about human behavior
Assess educational interventions
Make data-driven business decisions
p-value is the probability H₀ is true
p-value assumes H₀ is true — it's the probability of data given H₀, not the probability of H₀
p-value = P(data | H₀), not P(H₀ | data)
p > 0.05 means H₀ is true
Non-significance doesn't prove H₀ — just insufficient evidence to reject
Failure to reject H₀ ≠ accepting H₀
Using one-tailed test after seeing data direction
Direction must be specified before data collection
One-tailed tests require pre-registered hypotheses
Ignoring effect size for significant results
Statistical significance ≠ practical importance
Always report effect size (Cohen's d, r², etc.) alongside p-values
Alpha (α) is the predetermined threshold (usually 0.05). The p-value is calculated from data. If p < α, results are statistically significant. α controls Type I error rate; p-value measures evidence strength against H₀.
Type I error (false positive): Rejecting true H₀. Probability = α. Type II error (false negative): Failing to reject false H₀. Probability = β. Power = 1-β (probability of detecting true effect).
Larger samples make p-values smaller for the same effect size, as they provide more precise estimates. Very large samples can make trivial effects statistically significant. Always consider effect size, not just p-value.
Statistical significance (p < 0.05) indicates results unlikely due to chance. Practical significance (effect size) indicates real-world importance. A tiny effect can be statistically significant with large samples but practically meaningless.
95% CI shows range of plausible population values. If CI excludes null value (0 for difference, 1 for ratio), p < 0.05. CIs provide more information than p-values alone, showing precision and direction of effects.
When conducting multiple comparisons to control family-wise Type I error rate. Adjusted α = original α / number of tests. Example: 10 tests at α=0.05 → adjusted α=0.005. More conservative but increases Type II error risk.
P-hacking is manipulating analysis (e.g., adding/removing subjects, testing multiple outcomes) until p < 0.05. It inflates false positives and undermines reproducibility. Pre-register analyses and report all tested hypotheses.
Report exact p-values: p = .023, not p < .05. For very small p-values: p < .001. Include test statistic with df: t(28) = 2.15, p = .040. Report effect size: Cohen's d = 0.45. Never say 'marginally significant' (p < .10) without justification.
1890-1962
Introduced p-value concept (1925)
Set α = 0.05 as convention, also developed ANOVA
1894-1981
Formalized hypothesis testing framework
Developed confidence intervals with Egon Pearson
1876-1937
Developed t-distribution for small samples
Worked for Guinness Brewery, published as 'Student'
Did You Know?Ronald Fisher introduced the p-value concept in his 1925 book "Statistical Methods for Research Workers." He suggested p < 0.05 as a convenient threshold but warned against treating it as a rigid rule. The American Statistical Association (ASA) released a statement in 2016 clarifying p-value misconceptions and promoting better statistical practice.
Significant Result:
"The difference between groups was statistically significant, t(28) = 2.15, p = .040, Cohen's d = 0.45 [95% CI: 0.12, 0.78]."Non-Significant Result:
"No significant difference was found between conditions, t(28) = 0.85, p = .402."Very Small P-Value:
"The treatment effect was highly significant, t(28) = 4.52, p < .001."Key Rules:Report exact p-values (p = .023), not p < .05. For p < .001, report as p < .001. Include test statistic with degrees of freedom. Always report effect sizes and confidence intervals.
P-value calculator and complete guide. Calculate p-values for Z-tests, T-tests, and Chi-square tests. Learn hypothesis testing, significance levels (α = 0.05), one-tailed vs two-tailed tests, effect sizes, and common misconceptions. Essential for researchers, data scientists, and statisticians.