What is a real-life example of Pearson correlation?

A classic real-life example is the relationship between study hours and exam scores. Students who study more hours typically earn higher scores, producing a strong positive Pearson correlation (r ≈ 0.85–0.95). Another example is advertising spend vs. sales revenue: as ad spend increases, sales tend to rise, often showing r ≈ 0.80–0.90.

What does a Pearson correlation of 0.8 mean?

A Pearson correlation of r = 0.8 means a strong positive linear relationship between two variables. About 64% of the variance in one variable is explained by the other (r² = 0.64). When one variable increases, the other tends to increase as well, and this pattern is consistent.

What is considered a strong Pearson correlation?

Using Cohen's (1988) conventions: |r| ≥ 0.50 is strong, |r| between 0.30 and 0.49 is moderate, and |r| between 0.10 and 0.29 is weak. However, these thresholds vary by field. In psychology r = 0.30 may be considered meaningful; in physics or engineering, researchers expect r > 0.90.

Can Pearson correlation be negative?

Yes. A negative Pearson correlation means that as one variable increases, the other tends to decrease. For example, the correlation between exercise frequency and resting heart rate is negative (r ≈ −0.65 to −0.80): people who exercise more tend to have lower resting heart rates.

What is the difference between Pearson and Spearman correlation?

Pearson correlation measures the strength of a linear relationship between two continuous, normally distributed variables. Spearman correlation measures the strength of a monotonic relationship using ranked data and makes no assumption of normality. Use Spearman when your data has outliers, is ordinal, or is not normally distributed.

Pearson Correlation: 10 Real-Life Examples Explained (2026)

Q: What is Pearson correlation?

Pearson correlation (r) is a statistical measure that quantifies the strength and direction of the linear relationship between two continuous variables. Its value ranges from −1 to +1, where +1 is a perfect positive relationship, −1 is a perfect negative relationship, and 0 means no linear relationship.

What Is Pearson Correlation? (Definition)

Definition — Pearson Correlation Coefficient

The Pearson correlation coefficient (r) is a numerical measure of the strength and direction of the linear relationship between two continuous variables. Developed by Karl Pearson in 1895, it ranges from −1 to +1. A value of +1 means a perfect positive linear relationship; −1 means a perfect negative linear relationship; 0 means no linear relationship.

Range: −1 ≤ r ≤ +1

The intuition is simple. Scatter both variables on a plot. If the points form a straight line sloping upward, r is close to +1. If they form a line sloping downward, r is close to −1. If the cloud of points has no clear direction, r is near zero. The formula turns that visual impression into a single number you can report, test for significance, and compare across studies.

Pearson correlation was developed as part of Francis Galton's work on regression to the mean in the 1880s and mathematically formalized by Karl Pearson in 1895 in the paper "Notes on regression and inheritance in the case of two parents." Ronald Fisher later derived the sampling distribution of r and the significance test used today. Background on these ideas is in the statistics and probability section of Statistics Fundamentals.

⚡ Quick Reference — Pearson Correlation Key Facts

Symbol: r (sample) or ρ (population, "rho")
Range: −1 to +1 — values outside this range indicate a calculation error
Direction: Positive r = variables move together; Negative r = variables move in opposite directions
Magnitude: Closer to ±1 = stronger relationship; Closer to 0 = weaker relationship
R-squared (r²): The proportion of variance in one variable explained by the other
Only measures linear relationships: A curved or U-shaped relationship can give r ≈ 0 even when variables are strongly related
Requires continuous data: Both variables must be interval or ratio scale

The Pearson Correlation Formula

There are two equivalent versions of the formula — the definitional form using deviations from the mean, and the computational form using raw sums. Both give the same answer; the definitional form makes the concept clearest.

Definitional Formula

Pearson Correlation Coefficient — Definitional Form

r = Σ[(xᵢ − x̄)(yᵢ − ȳ)] / √[Σ(xᵢ − x̄)² · Σ(yᵢ − ȳ)²]

xᵢ = each x value ȳ = mean of y x̄ = mean of x yᵢ = each y value n = number of pairs Σ = sum over all pairs

In words: compute how much each observation deviates from its variable's mean, multiply those deviations together for each pair, sum across all pairs (the numerator), then divide by the square root of the product of the total squared deviations in x and y (the denominator). The denominator rescales the result to lie between −1 and +1 regardless of the original units.

Computational (Raw Score) Formula

Pearson Correlation Coefficient — Computational Form

r = [nΣxy − (Σx)(Σy)] / √{[nΣx² − (Σx)²][nΣy² − (Σy)²]}

n = sample size Σxy = sum of x×y products Σx, Σy = sums of each variable Σx², Σy² = sum of squared values

The computational form is faster when doing arithmetic by hand because you do not need to subtract the mean at each step. Both are formally derived in the NIST Engineering Statistics Handbook and in most introductory statistics textbooks such as Moore, McCabe, and Craig's Introduction to the Practice of Statistics.

How to Interpret Pearson's r

A number alone is incomplete. Interpretation requires three things: the sign (direction), the magnitude (strength), and context about the field. A correlation of 0.30 means something different in clinical psychology, where individual variation is enormous, than in precision manufacturing, where tight tolerances are the norm.

The most widely cited framework comes from Jacob Cohen's 1988 book Statistical Power Analysis for the Behavioral Sciences. His benchmarks apply primarily to the social and behavioral sciences. The table below shows those thresholds alongside what each range means in practice.

Strong+ r = 0.70 to 1.00 Strong positive linear relationship. Points cluster tightly around an upward-sloping line.

Moderate+ r = 0.40 to 0.69 Moderate positive relationship. A clear upward trend with noticeable scatter.

Weak+ r = 0.10 to 0.39 Weak positive relationship. A slight upward tendency but lots of scatter.

None r = −0.09 to 0.09 No linear relationship. The scatter plot shows no clear direction.

Weak− r = −0.39 to −0.10 Weak negative relationship. A slight downward tendency with lots of scatter.

Moderate− r = −0.69 to −0.40 Moderate negative relationship. A clear downward trend with noticeable scatter.

Strong− r = −1.00 to −0.70 Strong negative linear relationship. Points cluster tightly around a downward-sloping line.

Source: Cohen, J. (1988). Statistical Power Analysis for the Behavioral Sciences (2nd ed.). Hillsdale, NJ: Lawrence Erlbaum. Thresholds at 0.10, 0.30, and 0.50 for weak, moderate, and strong effects.

Understanding R-squared (r²)

Squaring the correlation coefficient gives r², called the coefficient of determination. This is the proportion of variance in one variable that is statistically associated with variance in the other. For example, r = 0.70 gives r² = 0.49, meaning 49% of the variation in y is accounted for by its linear relationship with x.

r = 0.9

r² = 0.81 → 81% shared variance

r = 0.7

r² = 0.49 → 49% shared variance

r = 0.5

r² = 0.25 → 25% shared variance

r = 0.3

r² = 0.09 → 9% shared variance

Positive Correlation Examples

A positive correlation means both variables tend to increase together. The following examples use original datasets computed specifically for this page. Each dataset contains ten paired observations — realistic figures drawn from the relevant research literature.

Example 1 — Study Hours vs. Exam Scores

Students who spend more time studying typically earn higher exam scores. This is one of the most studied correlations in educational psychology. The dataset below represents a class of ten students.

Worked Example 1 — Education

Original Dataset: Study Hours (x) vs. Exam Score % (y)

Student	Study Hours (x)	Exam Score % (y)	x − x̄	y − ȳ	(x−x̄)(y−ȳ)	(x−x̄)²	(y−ȳ)²
A	2	55	−4	−20	80	16	400
B	3	62	−3	−13	39	9	169
C	4	68	−2	−7	14	4	49
D	4	71	−2	−4	8	4	16
E	5	73	−1	−2	2	1	4
F	6	77	0	2	0	0	4
G	7	82	1	7	7	1	49
H	8	85	2	10	20	4	100
I	9	90	3	15	45	9	225
J	12	87	6	12	72	36	144
Σ	60	750	—	—	287	84	1160

x̄ = 60/10 = 6.0 hours | ȳ = 750/10 = 75.0%

Sum of cross-products (numerator): Σ[(xᵢ − x̄)(yᵢ − ȳ)] = 80 + 39 + 14 + 8 + 2 + 0 + 7 + 20 + 45 + 72 = 287

Sum of squared deviations for x: Σ(xᵢ − x̄)² = 16 + 9 + 4 + 4 + 1 + 0 + 1 + 4 + 9 + 36 = 84

Sum of squared deviations for y: Σ(yᵢ − ȳ)² = 400 + 169 + 49 + 16 + 4 + 4 + 49 + 100 + 225 + 144 = 1160

Apply the formula:
r = 287 / √(84 × 1160) = 287 / √97440 = 287 / 312.2 = r = 0.92

✅ Interpretation: r = 0.92 indicates a strong positive linear relationship. r² = 0.846, meaning about 85% of the variance in exam scores is explained by study hours. This is consistent with research findings: in a 2020 meta-analysis published in Review of Educational Research, study time was among the strongest predictors of academic performance.

Scatter Plot — Study Hours vs. Exam Score (r = 0.92)

Example 2 — Height vs. Weight

Taller people tend to weigh more. This relationship appears across populations and is a standard example in introductory statistics courses. Data below uses adult measurements in metric units.

Worked Example 2 — Anthropometry

Original Dataset: Height in cm (x) vs. Weight in kg (y)

Person	Height cm (x)	Weight kg (y)	x²	y²	xy
1	155	52	24025	2704	8060
2	160	58	25600	3364	9280
3	163	61	26569	3721	9943
4	167	65	27889	4225	10855
5	170	68	28900	4624	11560
6	173	72	29929	5184	12456
7	175	74	30625	5476	12950
8	178	80	31684	6400	14240
9	182	85	33124	7225	15470
10	188	92	35344	8464	17296
Σ	1711	707	293689	51387	122110

Numerator: nΣxy − (Σx)(Σy) = 10×122110 − 1711×707 = 1221100 − 1209677 = 11423

Denominator part 1: nΣx² − (Σx)² = 10×293689 − 1711² = 2936890 − 2927521 = 9369

Denominator part 2: nΣy² − (Σy)² = 10×51387 − 707² = 513870 − 499849 = 14021

Calculate r: r = 11423 / √(9369 × 14021) = 11423 / √131346549 = 11423 / 11461.1 = r = 0.997

✅ Interpretation: r = 0.997 is an extremely strong positive correlation. Height and weight are almost perfectly linearly related in this dataset. In practice, population-level correlations between adult height and weight typically range from r = 0.60 to r = 0.70, with greater scatter from differences in body composition, age, and sex — the high value here reflects a small, homogeneous illustrative sample.

Example 3 — Advertising Spend vs. Sales Revenue

Businesses routinely examine whether advertising budgets translate into higher sales. A moderate-to-strong positive correlation is typical, though the relationship depends on the industry and advertising medium.

Worked Example 3 — Business Analytics

Monthly Ad Spend $'000 (x) vs. Sales Revenue $'000 (y) — 10 Months

Month	Ad Spend (x)	Sales (y)	x − x̄	y − ȳ	(x−x̄)(y−ȳ)	(x−x̄)²	(y−ȳ)²
Jan	10	80	−8	−40	320	64	1600
Feb	12	95	−6	−25	150	36	625
Mar	15	105	−3	−15	45	9	225
Apr	18	118	0	−2	0	0	4
May	20	115	2	−5	−10	4	25
Jun	22	132	4	12	48	16	144
Jul	25	140	7	20	140	49	400
Aug	24	138	6	18	108	36	324
Sep	19	125	1	5	5	1	25
Oct	15	152	−3	32	−96	9	1024
Σ	180	1200	—	—	710	224	4396

x̄ = 18 | ȳ = 120 | r = 710 / √(224 × 4396) = 710 / √984704 = 710 / 992.3 = r = 0.716

✅ Interpretation: r = 0.716 indicates a strong positive relationship between advertising spend and sales. October is an outlier — high sales despite lower ad spend, possibly due to seasonal demand — which pulls r below 0.90. This illustrates why scatter plots should always accompany the correlation coefficient.

Example 4 — Household Income vs. Monthly Spending

Higher-income households tend to spend more each month. This relationship is studied by economists and market researchers and typically shows a strong positive correlation, though it flattens at higher income levels (suggesting the linear model is approximate).

Worked Example 4 — Economics

Annual Household Income $'000 (x) vs. Monthly Spending $ (y)

#	Income $k (x)	Spending $ (y)	xy	x²	y²
1	35	1800	63000	1225	3240000
2	45	2100	94500	2025	4410000
3	55	2400	132000	3025	5760000
4	65	2600	169000	4225	6760000
5	75	2900	217500	5625	8410000
6	85	3100	263500	7225	9610000
7	100	3400	340000	10000	11560000
8	120	3700	444000	14400	13690000
9	150	4200	630000	22500	17640000
10	200	5000	1000000	40000	25000000
Σ	930	31200	3353500	110250	106080000

r = (10×3353500 − 930×31200) / √[(10×110250 − 930²)(10×106080000 − 31200²)]
= (33535000 − 29016000) / √[(1102500 − 864900)(1060800000 − 973440000)]
= 4519000 / √(237600 × 87360000) = 4519000 / √20758336000000 = 4519000 / 4556130 ≈ r = 0.99

✅ Interpretation: r = 0.99 in this dataset reflects an almost perfectly linear relationship. Income and spending are very tightly coupled here. Real survey data from sources such as the Bureau of Labor Statistics Consumer Expenditure Survey shows r values typically in the range 0.85–0.95, with more scatter at higher income levels where savings rates increase.

Negative Correlation Examples

A negative correlation means one variable tends to decrease as the other increases. The data points slope downward from left to right on a scatter plot. The following examples illustrate common negative correlations across different fields.

Example 5 — Exercise Frequency vs. Resting Heart Rate

People who exercise more consistently tend to have lower resting heart rates, because a conditioned heart pumps blood more efficiently. This is a well-established finding in exercise physiology.

Worked Example 5 — Healthcare / Exercise Physiology

Weekly Exercise Sessions (x) vs. Resting Heart Rate bpm (y)

Person	Sessions/wk (x)	Resting HR bpm (y)	x − x̄	y − ȳ	(x−x̄)(y−ȳ)	(x−x̄)²	(y−ȳ)²
1	0	88	−3.5	14.5	−50.75	12.25	210.25
2	1	82	−2.5	8.5	−21.25	6.25	72.25
3	1	84	−2.5	10.5	−26.25	6.25	110.25
4	2	78	−1.5	4.5	−6.75	2.25	20.25
5	3	74	−0.5	0.5	−0.25	0.25	0.25
6	4	71	0.5	−2.5	−1.25	0.25	6.25
7	4	68	0.5	−5.5	−2.75	0.25	30.25
8	5	65	1.5	−8.5	−12.75	2.25	72.25
9	6	60	2.5	−13.5	−33.75	6.25	182.25
10	9	65	5.5	−8.5	−46.75	30.25	72.25
Σ	35	735	—	—	−202.5	66.5	776.5

x̄ = 3.5 | ȳ = 73.5 | r = −202.5 / √(66.5 × 776.5) = −202.5 / √51636.25 = −202.5 / 227.2 = r = −0.89

✅ Interpretation: r = −0.89 indicates a strong negative linear relationship. As exercise sessions per week increase, resting heart rate tends to decrease. The American Heart Association defines a healthy resting HR as 60–100 bpm; endurance athletes can have resting HR as low as 40 bpm, consistent with this pattern.

Scatter Plot — Exercise Sessions vs. Resting Heart Rate (r = −0.89)

Example 6 — Temperature vs. Heating Bills

As outdoor temperatures rise, households use less heating, so energy bills fall. This is a straightforward negative correlation observed across utility data in temperate climates.

Worked Example 6 — Environmental / Consumer Data

Average Monthly Temperature °C (x) vs. Heating Bill $ (y) — 10 Months

Month	Temp °C (x)	Heating $ (y)	xy	x²	y²
Jan	−5	280	−1400	25	78400
Feb	−2	250	−500	4	62500
Mar	5	190	950	25	36100
Apr	10	140	1400	100	19600
May	16	85	1360	256	7225
Jun	22	30	660	484	900
Jul	26	15	390	676	225
Aug	25	18	450	625	324
Sep	18	60	1080	324	3600
Oct	8	165	1320	64	27225
Σ	123	1233	5710	2583	236099

r = (10×5710 − 123×1233) / √[(10×2583 − 123²)(10×236099 − 1233²)]
= (57100 − 151659) / √[(25830 − 15129)(2360990 − 1520289)]
= −94559 / √(10701 × 840701) = −94559 / √8996240001 ≈ −94559 / 94849 ≈ r = −0.997

✅ Interpretation: r = −0.997 is an extremely strong negative correlation — nearly perfect. Temperature is an almost perfect linear predictor of heating costs over this range. This makes practical sense: heating demand responds directly and predictably to outdoor temperatures over a typical seasonal range.

Example 7 — Hours of Sleep vs. Cognitive Performance

Sleep deprivation impairs working memory, reaction time, and decision-making. Research consistently shows that people who sleep fewer hours perform worse on cognitive tasks. The dataset below uses a simulated study modeled on findings from the Journal of Sleep Research.

Worked Example 7 — Psychology / Neuroscience

Hours of Sleep (x) vs. Cognitive Error Rate % (y)

Note: Higher y = more errors = worse performance. So more sleep → fewer errors = negative correlation.

#	Sleep hrs (x)	Error Rate % (y)	x − x̄	y − ȳ	(x−x̄)(y−ȳ)	(x−x̄)²	(y−ȳ)²
1	4	42	−3	18.3	−54.9	9	334.89
2	4.5	38	−2.5	14.3	−35.75	6.25	204.49
3	5	33	−2	9.3	−18.6	4	86.49
4	5.5	30	−1.5	6.3	−9.45	2.25	39.69
5	6	26	−1	2.3	−2.3	1	5.29
6	6.5	22	−0.5	−1.7	0.85	0.25	2.89
7	7	20	0	−3.7	0	0	13.69
8	7.5	17	0.5	−6.7	−3.35	0.25	44.89
9	8	14	1	−9.7	−9.7	1	94.09
10	9	11	2	−12.7	−25.4	4	161.29
Σ	63	253	—	—	−158.6	28	987.7

x̄ = 6.3 | ȳ = 25.3 | r = −158.6 / √(28 × 987.7) = −158.6 / √27655.6 = −158.6 / 166.3 ≈ r = −0.954

✅ Interpretation: r = −0.954 is a strong negative correlation. More sleep hours are associated with fewer cognitive errors. This direction is consistent with an extensive body of sleep science research. The relationship does plateau around 8–9 hours: sleeping 10 hours does not necessarily outperform 8, which is why a linear model may be approximate at the extremes.

Zero (No Linear) Correlation Examples

A correlation near zero means there is no consistent linear relationship between the two variables. Zero correlation does not mean the variables are unrelated — they could have a strong curved (non-linear) relationship that Pearson's r fails to detect.

Example 8 — Shoe Size vs. Annual Salary

Worked Example 8 — Spurious / Zero Correlation

Adult Shoe Size (x) vs. Annual Salary $'000 (y) — n = 10

#	Shoe Size (x)	Salary $k (y)	x − x̄	y − ȳ	(x−x̄)(y−ȳ)	(x−x̄)²	(y−ȳ)²
1	6	52	−2	−11	22	4	121
2	6.5	80	−1.5	17	−25.5	2.25	289
3	7	48	−1	−15	15	1	225
4	7.5	67	−0.5	4	−2	0.25	16
5	8	55	0	−8	0	0	64
6	8	72	0	9	0	0	81
7	8.5	63	0.5	0	0	0.25	0
8	9	45	1	−18	−18	1	324
9	9.5	90	1.5	27	40.5	2.25	729
10	10	58	2	−5	−10	4	25
Σ	80	630	—	—	22	15	1874

x̄ = 8.0 | ȳ = 63.0 | r = 22 / √(15 × 1874) = 22 / √28110 = 22 / 167.7 ≈ r = 0.13

✅ Interpretation: r = 0.13 is negligible — essentially zero. Shoe size has no meaningful linear relationship with salary. This example illustrates why it is important to choose variables with a plausible causal or theoretical connection before running a correlation analysis.

More Real-Life Pearson Correlation Examples

The following examples are reported as summary results with interpretations rather than full worked arithmetic, to cover a broader set of fields. Each describes a common application and gives a typical r value from the research literature.

Example 9 — IQ Score vs. Academic GPA (Psychology)

Case Study — Educational Psychology

IQ and Academic Achievement

Researchers studying cognitive predictors of academic performance measure the correlation between standardized IQ scores and cumulative GPA across a sample of college students. Meta-analyses in educational psychology consistently find correlations in this range.

r = 0.50

Typical finding in educational psychology literature

r² = 0.25

About 25% of GPA variance explained by IQ

Moderate+

Strength classification (Cohen, 1988)

Interpretation: A moderate positive correlation. IQ is one predictor of academic performance, but 75% of GPA variance is explained by other factors — study habits, motivation, subject interest, instructional quality, and socioeconomic factors. This shows why a correlation of 0.50, though meaningful, does not allow precise individual-level prediction.

Reference: Deary, I.J., Strand, S., Smith, P., & Fernandes, C. (2007). Intelligence and educational achievement. Intelligence, 35(1), 13–21. Reports r ≈ 0.61 between cognitive ability at age 11 and GCSE examination scores at age 16.

Example 10 — Website Visitors vs. Conversions (Digital Marketing)

Case Study — Digital Marketing

Traffic and Conversion Rate

A digital marketing team tracks monthly website visitors and the number of conversions (purchases, sign-ups) over 12 months. They want to know whether more traffic reliably produces more conversions, or whether conversion rate fluctuates independently.

r = 0.78

Strong positive correlation (traffic → conversions)

r² = 0.61

61% of conversion variance explained by traffic

Strong+

More traffic generally means more conversions

Interpretation: A strong positive correlation. Higher traffic volumes tend to produce more conversions, but the relationship is imperfect — campaign quality, landing page effectiveness, and seasonality all affect whether a visitor converts. Using this correlation, the team can forecast conversion volume from planned traffic targets, with appropriate uncertainty bounds from a confidence interval.

How to Calculate Pearson Correlation Step by Step

The manual calculation uses the definitional formula and follows these steps consistently. The same procedure applies regardless of the field or variable names.

Collect Paired Data

Each observation must consist of two measurements taken from the same unit (person, time period, location). Record n paired (xᵢ, yᵢ) values. Both variables must be continuous (interval or ratio scale). Check for obvious data entry errors before computing.

Compute the Means (x̄ and ȳ)

Sum all x values and divide by n to get x̄. Repeat for y to get ȳ. These are the reference points from which deviations are measured. The formula for the mean is covered in the mean guide.

Compute the Deviations

For each observation, calculate (xᵢ − x̄) and (yᵢ − ȳ). These are the deviations from the mean. Verify that the sum of each deviation column equals zero — this is a useful arithmetic check.

Compute the Products and Squares

For each pair, calculate the cross-product (xᵢ − x̄)(yᵢ − ȳ) and the squared deviations (xᵢ − x̄)² and (yᵢ − ȳ)². Then sum each of those three columns to get Σ[(xᵢ − x̄)(yᵢ − ȳ)], Σ(xᵢ − x̄)², and Σ(yᵢ − ȳ)².

Apply the Formula

r = Σ[(xᵢ − x̄)(yᵢ − ȳ)] / √[Σ(xᵢ − x̄)² × Σ(yᵢ − ȳ)²]. The result must be between −1 and +1. If yours falls outside this range, recheck your arithmetic — this cannot happen with a correct calculation.

Test for Statistical Significance

Use the t-statistic: t = r√(n − 2) / √(1 − r²), with n − 2 degrees of freedom. Compare to a critical value from the t-distribution table. At α = 0.05 with n = 10, the critical value is approximately ±2.306. If |t| exceeds this, the correlation is statistically significant. Full critical values are in the Pearson correlation critical values table.

Interpret the Result in Context

State the direction (positive or negative), magnitude (using Cohen's benchmarks), the r² value, whether the result is statistically significant, and what this means for the research question. Correlation does not establish causation — see the correlation vs. causation guide.

Calculating in R, Python, Excel, and SPSS

🔬

R

cor(x, y, method = "pearson")
Returns r directly. For significance: cor.test(x, y)

🐍

Python (SciPy)

from scipy import stats r, p = stats.pearsonr(x, y)
Returns r and two-tailed p-value.

📊

Excel

=CORREL(array1, array2)
Or use Data → Data Analysis → Correlation for a full matrix.

📈

SPSS

Analyze → Correlate → Bivariate. Select Pearson. Output includes r, n, and two-tailed significance for each pair.

Pearson Correlation Assumptions

Pearson's r is only appropriate when specific conditions are met. Violating these conditions can produce a correlation coefficient that misrepresents the actual relationship between variables.

⚠️

Check Assumptions Before Calculating

Pearson's r can be computed for any two numeric variables, but the result is only meaningful when the assumptions below are satisfied. Always plot your data first — a scatter plot reveals violations that the formula cannot detect.

Assumption	What It Means	How to Check
Continuous data	Both variables are interval or ratio scale — not ordinal, nominal, or binary	Check measurement scale of each variable
Linear relationship	The relationship between x and y is linear (straight-line), not curved	Inspect the scatter plot for a curved pattern
Bivariate normality	Both variables are approximately normally distributed (especially for hypothesis testing)	Histograms, Q-Q plots, normality tests
No extreme outliers	Outliers can inflate or deflate r substantially — a single outlier can change r by 0.3 or more	Scatter plot, outlier detection
Homoscedasticity	The spread of y values should be roughly constant across all values of x	Scatter plot — look for funnel or fan patterns
Independent observations	Each data pair is independent — no repeated measures or time-series autocorrelation	Study design; use paired t-test or time-series methods if needed

When these assumptions are violated, the Spearman rank correlation is often a better choice. Spearman makes no assumption of normality, is less sensitive to outliers, and works with ordinal data — at the cost of slightly less statistical power when the Pearson assumptions are met.

Comparison Tables

Pearson vs. Spearman Correlation

Feature	Pearson (r)	Spearman (ρ)
What it measures	Strength of linear relationship	Strength of monotonic relationship
Data type required	Continuous (interval/ratio)	Ordinal, interval, or ratio
Normality required	Yes (for inference)	No
Sensitive to outliers	Yes — substantially	Less sensitive (uses ranks)
Works with non-linear relationships	No — misses curves	Yes, if monotonic
Formula	Based on means and standard deviations	Based on rank differences
Use when	Continuous, normal data, linear trend expected	Ordinal data, outliers present, or normality violated
Statistical power	Higher when assumptions are met	Slightly lower under ideal conditions

Correlation vs. Causation

🚫

Correlation Does Not Imply Causation

A significant Pearson correlation tells you two variables are associated — not that one causes the other. Ice cream sales and drowning rates are positively correlated (r ≈ 0.80 in summer months) because both are driven by a third variable: hot weather. Always consider confounding variables, reverse causation, and chance findings before drawing causal conclusions. See the full guide on correlation vs. causation.

Aspect	Correlation	Causation
What it shows	Statistical association between two variables	One variable directly produces change in another
Evidence strength	Can be established from observational data	Requires controlled experiments or strong causal theory
Confounding	May be driven by a third variable	Controlled experiments eliminate confounders
Direction	Bidirectional — does not specify which causes which	Directional — cause precedes effect
Example	r = 0.85 between shoe size and reading ability in children (both caused by age)	Smoking causes lung cancer (established via randomized evidence and mechanism)

The CORRELATE Framework

The CORRELATE framework provides a structured procedure for applying Pearson correlation correctly in any analytical context. It is designed to prevent the most common mistakes: computing r without checking assumptions, reporting r without p-values, and confusing correlation with causation.

🧭

Original Framework — CORRELATE

An 8-step protocol developed by Statistics Fundamentals to standardize how Pearson correlation is applied and reported across fields.

Collect Paired Data

Ensure you have n paired continuous measurements. Both variables must come from the same unit of observation. Define the research question before collecting data to avoid data dredging.

Observe the Scatter Plot

Plot y against x before computing any statistic. Look for linearity, outliers, heteroscedasticity (funnel shapes), and curvature. A scatter plot often tells you more than the correlation coefficient alone.

Review Assumptions

Confirm continuous scale, approximate normality, no extreme outliers, and homoscedasticity. If assumptions are violated, consider Spearman correlation or data transformation before proceeding.

Run the Pearson Correlation

Apply the formula manually using the worked-example method above, or use statistical software (R, Python, SPSS, Excel). Always record n alongside r — a correlation of 0.80 from n = 5 is far less reliable than from n = 500.

Evaluate Strength

Classify r using Cohen's (1988) benchmarks: |r| ≥ 0.50 strong, 0.30–0.49 moderate, 0.10–0.29 weak. Compute r² to understand how much shared variance the correlation explains.

Look for Significance

Test whether r differs significantly from zero using t = r√(n − 2)/√(1 − r²) with df = n − 2. Report the p-value and confidence interval for r, not just the correlation alone. Use the Pearson correlation table for critical values.

Apply Interpretation

State the direction (positive or negative), magnitude (weak/moderate/strong), r², significance, and what the result means for your specific question. Reference field-specific norms — what is "strong" in psychology may be "weak" in physics.

Translate to Real-World Meaning

Convert the statistical finding into language accessible to a non-specialist audience. Example: "Students who studied one additional hour per day scored about 4 percentage points higher on average" is clearer than "r = 0.92."

Explain Limitations

State what the correlation does not tell you: it does not establish causation, it only measures linear relationships, it is affected by outliers, and it may not generalize beyond the sample studied. Mention any violated assumptions.

When to Use (and Not Use) Pearson Correlation

Decision Guide — Is Pearson Correlation the Right Choice?

Are both variables continuous (interval or ratio)?

→ No →

Use Spearman (ordinal data) or Point Biserial (one binary variable)

Does the scatter plot show a roughly linear pattern?

→ No →

Consider Spearman or a nonlinear model — Pearson r will understate the relationship

Are both variables approximately normally distributed?

→ No →

Pearson r is still calculable; significance test is unreliable. Use Spearman for safer inference

Are there extreme outliers present?

→ Yes →

Remove or examine outliers first — a single outlier can change r dramatically

All conditions met — proceed with Pearson correlation

→ Yes →

Calculate r, test significance, report with n and p-value

Common Mistakes with Pearson Correlation

Common Mistake	What Is Wrong	Correct Approach
"r = 0.85, so one variable causes the other"	Confuses association with causation	Report as association only; use experimental design or causal inference methods to establish causation
"r = 0 means the variables are unrelated"	Zero linear correlation does not rule out a strong curved relationship	Plot the data — Anscombe's quartet famously shows four datasets with identical r but very different patterns
"A significant p-value means a large effect"	With large n, even r = 0.05 can be statistically significant	Always report r and r² alongside the p-value; significance and practical importance are not the same
Using Pearson with ordinal data (e.g., Likert scales)	Ordinal data does not have equal intervals; Pearson assumes continuous scale	Use Spearman correlation or polychoric correlation for ordinal variables
Not checking for outliers before computing r	A single outlier can inflate or deflate r by 0.2–0.5 in small samples	Always plot the data first; report results with and without outliers when they are present
Reporting only r without sample size or p-value	r = 0.70 from n = 5 is not reliable; from n = 500 it is	Always report: r, n, p-value, and ideally a 95% confidence interval for r

Entity and Formula Glossary

Term	Symbol	Definition
Pearson Correlation	`r`	A measure of the strength and direction of the linear relationship between two continuous variables. Ranges from −1 to +1.
Correlation Coefficient	`r or ρ`	Generic term for a standardized measure of association; Pearson's r is the most common form for continuous data.
Scatter Plot	—	A graph plotting paired data as points, with x on the horizontal axis and y on the vertical axis. Used to visualize correlation direction and linearity.
Positive Correlation	`r > 0`	Both variables tend to increase together. Example: study hours and exam scores.
Negative Correlation	`r < 0`	One variable increases as the other decreases. Example: exercise frequency and resting heart rate.
Zero Correlation	`r ≈ 0`	No consistent linear relationship. Example: shoe size and salary.
Covariance	`Cov(x,y)`	The unnormalized version of Pearson correlation: Cov(x,y) = Σ[(xᵢ−x̄)(yᵢ−ȳ)]/(n−1). Pearson r normalizes this by the standard deviations.
R-squared	`r²`	The coefficient of determination. The proportion of variance in y explained by its linear relationship with x.
Outlier	—	A data point that falls far from the general pattern. Can substantially change the value of r, especially in small samples.
p-value	`p`	The probability of observing a correlation as extreme as r by chance if the true population correlation is zero. See the p-values guide.
Statistical Significance	`p < α`	The correlation is unlikely to have occurred by chance alone at the chosen significance level α (usually 0.05).
Confidence Interval	—	A range of plausible values for the true population correlation ρ, computed via Fisher's z-transformation. A 95% CI means 95% of such intervals contain the true ρ. See the confidence intervals guide.
Linear Relationship	—	A relationship that can be represented by a straight line on a scatter plot. Pearson r measures only this type of relationship.
Correlation Matrix	—	A table showing Pearson r values for all pairs of variables in a dataset. Used in multivariate analysis and factor analysis.
Karl Pearson	—	British statistician (1857–1936) who mathematically formalized the correlation coefficient in 1895. Also developed the chi-square goodness-of-fit test.
Spearman Correlation	`ρ`	A rank-based alternative to Pearson correlation, suitable for ordinal data or when Pearson assumptions are violated. See the Spearman correlation guide.

Pearson Correlation Calculator

Enter paired x and y values below, one pair per line separated by a comma (e.g., 5, 72). The calculator computes r, r², the t-statistic, and the interpretation.

Interactive Pearson Correlation Calculator

X Values (one per line, or comma-separated on same line as Y)

r = —

Real-World Applications of Pearson Correlation

Pearson correlation appears across virtually every quantitative field. Below are the most common application areas, with the types of variable pairs typically studied and the correlations found in the literature.

🎓

Education Research

Study time vs. grades, IQ vs. GPA, class size vs. test scores, teacher experience vs. student performance. Typical r ranges: 0.30–0.70.

🧠

Psychology

Intelligence measures vs. job performance, anxiety scores vs. academic achievement, parental income vs. child development outcomes. Correlations 0.20–0.60 are common.

🏥

Healthcare

Body mass index vs. blood pressure, age vs. bone density, diet quality vs. cholesterol. Clinical correlations often range 0.40–0.80.

💼

Business Analytics

Ad spend vs. revenue, customer satisfaction vs. repeat purchases, price vs. demand. Marketing correlations vary widely, typically 0.50–0.90 for controlled analyses.

🌱

Environmental Science

CO₂ concentration vs. temperature, rainfall vs. crop yield, pollution levels vs. respiratory disease rates. Environmental correlations tend to be high (0.70–0.99) for physical relationships.

🏃

Sports Science

Training volume vs. race time, muscle mass vs. strength output, recovery hours vs. performance. Sports science correlations range 0.50–0.90 depending on the variables.

📊

Data Science / ML

Feature correlation analysis in preprocessing (identifying redundant features), measuring target variable relationships before model selection. See also the guide on simple linear regression, which extends correlation to prediction.

💹

Finance

Stock return correlations (for portfolio diversification), GDP vs. unemployment rate (Okun's Law), interest rates vs. bond prices. Financial correlations can shift substantially during market stress.

Pearson Correlation Cheat Sheet

Item	Detail
Formula (definitional)	r = Σ[(xᵢ−x̄)(yᵢ−ȳ)] / √[Σ(xᵢ−x̄)² · Σ(yᵢ−ȳ)²]
Formula (computational)	r = [nΣxy − (Σx)(Σy)] / √{[nΣx² − (Σx)²][nΣy² − (Σy)²]}
Range	−1 ≤ r ≤ +1
Perfect positive	r = +1.00
Perfect negative	r = −1.00
No linear relationship	r = 0
Strong (Cohen)	\|r\| ≥ 0.50
Moderate (Cohen)	\|r\| = 0.30–0.49
Weak (Cohen)	\|r\| = 0.10–0.29
R-squared	r² = proportion of shared variance
Significance test	t = r√(n−2) / √(1−r²), df = n−2
Degrees of freedom	df = n − 2
Critical values (α=0.05)	n=10: \|r\| > 0.632; n=20: \|r\| > 0.444; n=30: \|r\| > 0.361
R command	cor.test(x, y, method="pearson")
Python command	scipy.stats.pearsonr(x, y)
Excel command	=CORREL(array1, array2)
Alternative when assumptions fail	Spearman correlation (use with ordinal data or outliers)

Frequently Asked Questions

Pearson correlation is a statistic that measures the strength and direction of the linear relationship between two continuous variables. It ranges from −1 to +1. Values close to +1 indicate a strong positive relationship, values close to −1 indicate a strong negative relationship, and values near 0 indicate little or no linear relationship between the variables.

A common example is study time and exam scores, where students who study more generally earn higher grades, producing a positive correlation. Other examples include exercise frequency and resting heart rate (negative correlation) and outdoor temperature and heating costs (negative correlation).

A Pearson correlation coefficient of 0.8 indicates a strong positive linear relationship. As one variable increases, the other tends to increase as well. Squaring the correlation gives r² = 0.64, meaning approximately 64% of the variation in one variable is associated with the variation in the other.

Although interpretation depends on the field of study, many researchers consider an absolute correlation of 0.50 or greater to be strong. Values between 0.30 and 0.49 are often considered moderate, while values below 0.30 are generally viewed as weak.

Pearson correlation measures linear relationships between continuous variables and assumes approximately normally distributed data. Spearman rank correlation measures monotonic relationships using ranked values, making it more suitable for ordinal data, non-normal distributions, or datasets containing influential outliers. See the Spearman correlation guide for a detailed comparison.

Yes. A negative Pearson correlation indicates that the variables tend to move in opposite directions. As one variable increases, the other tends to decrease. A value of −1 represents a perfect negative linear relationship, while values closer to 0 indicate weaker negative relationships.

Pearson correlation should not be used when variables have a non-linear relationship, when data are ordinal rather than continuous, when extreme outliers strongly influence the results, or when observations are not independent. In these situations, Spearman correlation or Kendall's tau is often a better choice.

No. Correlation measures association, not cause and effect. Two variables may be correlated because of a third variable, reverse causation, or coincidence. Demonstrating causation requires additional evidence such as controlled experiments or well-designed causal studies. Learn more in the correlation vs. causation guide.

APA 7th edition recommends reporting Pearson correlation as r(df) = .XX, p = .XXX. For example: "There was a strong positive correlation between study hours and exam scores, r(28) = .92, p < .001." Report the degrees of freedom (n − 2), the correlation coefficient, and the p-value.

r², called the coefficient of determination, is the square of the Pearson correlation coefficient. It represents the proportion of variation in one variable that is explained by its linear relationship with the other variable. For example, if r = 0.70, then r² = 0.49, meaning 49% of the variation is explained by the relationship.

Sources and Further Reading

Pearson, K. (1895) — "Notes on regression and inheritance in the case of two parents." Proceedings of the Royal Society of London, 58, 240–242. Original paper introducing the correlation coefficient.
Cohen, J. (1988) — Statistical Power Analysis for the Behavioral Sciences (2nd ed.). Hillsdale, NJ: Lawrence Erlbaum Associates. Source of the widely used r = 0.10/0.30/0.50 benchmarks for weak, moderate, and strong correlations.
NIST/SEMATECH Engineering Statistics Handbook — Section 1.3.5.13: Pearson Correlation. itl.nist.gov
Moore, D.S., McCabe, G.P., & Craig, B.A. — Introduction to the Practice of Statistics (9th ed.). New York: W.H. Freeman. Standard undergraduate textbook covering Pearson correlation with worked examples.
Fisher, R.A. (1915) — "Frequency distribution of the values of the correlation coefficient in samples from an indefinitely large population." Biometrika, 10(4), 507–521. Fisher's z-transformation for confidence intervals around r.
SciPy Documentation — scipy.stats.pearsonr function. docs.scipy.org
American Statistical Association (2016) — "Statement on Statistical Significance and P-Values." The American Statistician, 70(2), 129–133. tandfonline.com
Penn State STAT 501 — Regression Methods, Chapter 1: Simple Linear Regression. online.stat.psu.edu