Google Sheets Handbook for the IB Maths IA

A step-by-step technical guide to statistical analysis and modelling for the Internal Assessment.

How to use this guide. Examples assume your raw data runs from A2:A101 (column A, rows 2-101). Adjust the range to match your own dataset. All formulas start with = and are typed directly into a cell.
Part 1

Descriptive Statistics & Data Representation

1 Summary Statistics Formulas

For any single-variable dataset in A2:A101, enter these formulas into separate summary cells. Build a labelled table in your IA to present them - don't just paste the Sheets output.

Core descriptive statistics
Mean (average)=AVERAGE(A2:A101)
Median=MEDIAN(A2:A101)
Mode=MODE(A2:A101)
Sample standard deviation=STDEV.S(A2:A101)
Population standard deviation=STDEV.P(A2:A101)
Lower quartile Q₁=QUARTILE(A2:A101, 1)
Upper quartile Q₃=QUARTILE(A2:A101, 3)
Interquartile range (IQR)=Cell_Q3 - Cell_Q1 (reference your Q₃ and Q₁ cells)
STDEV.P vs STDEV.S

STDEV.P divides by n and matches the standard deviation σ that your GDC gives for a data set. STDEV.S divides by n − 1 and estimates the standard deviation of a larger population from a sample. Decide which suits what your data represents, say which one you used and why, and use the same one throughout.

MODE limitations

=MODE returns only the first mode if multiple exist. For multimodal data use =MODE.MULT(A2:A101) (in Google Sheets the results spill into the cells below). If your data is continuous, mode is often not meaningful - focus on mean, median and standard deviation instead.

Criterion E tip - don't just list numbers. In your write-up, explicitly compare the mean and median. If they differ significantly, identify skewness in your data and explain how outliers pulled the mean away from the centre. For example: "The mean (47.3) exceeds the median (41.0), suggesting positive skew - a small number of high-value observations inflate the average." Explaining what a result means, not just reporting it, is how you show the understanding that Criterion E asks you to demonstrate.

2 Generating Histograms with Specified Bucket Sizes

Histograms display the frequency distribution of continuous data. Setting a uniform bucket (class interval) width makes the histogram easier to read and compare.

1
Highlight your data column

Select your continuous dataset including the header row (e.g. click the column A header to select the whole column, or drag from A1 to A101).

2
Insert the chart

Navigate to Insert → Chart. In the Chart Editor panel that appears on the right, open the Chart type dropdown and select Histogram chart.

3
Set a custom bucket size

Click the Customize tab inside the Chart Editor, expand the Histogram section, and change Bucket size from "Auto" to your preferred uniform class interval - for example 5.0 or 10.0. Equal-width intervals keep a frequency histogram simple to read; if you use unequal widths, plot frequency density instead.

4
Label axes and add a title

Still in Customize → Chart & axis titles: set the chart title, horizontal axis label (with units), and vertical axis label ("Frequency"). The IB guide gives graphs that are not labelled as an example of level 1 mathematical communication (Criterion B).

Chart Editor → Customize → Histogram
SettingValue
Bucket size5.0
Show item dividerschecked
Outlier percentile0% (show all data)
Part 2

Bivariate Data & Statistical Correlation

1 Plotting a Scatter Diagram with a Trendline & R² Value

To analyse the relationship between two variables, place your independent variable (X) in the left column and your dependent variable (Y) in the adjacent right column. Sheets reads the left column as x by default.

1
Highlight both data columns uniformly

Select from A1 to B101 (including headers). Both columns must be the same height - any missing values will distort the chart.

2
Insert a Scatter chart

Click Insert → Chart. In the Chart Editor, change the Chart type to Scatter chart. Sheets may auto-suggest a Line chart - always switch to Scatter for mathematical data so individual data points are shown correctly.

3
Add the trendline and display the equation

Switch to the Customize tab → expand Series → scroll down and tick Trendline. Change the Label dropdown to "Use equation". This outputs the exact regression formula (e.g. y = 2.34x + 5.17) directly onto the chart.

4
Show R²

In the same Series panel, tick "Show R²". The coefficient of determination appears on the chart automatically. Copy this value into your IA write-up and interpret it - do not just reproduce the chart without discussion.

5
Label axes and export

Under Chart & axis titles, add a descriptive chart title and both axis labels with units. Export via the three-dot menu → Download → PNG. Paste into your document with a figure caption below.

2 Calculating Pearson's Correlation Coefficient (r)

To mathematically justify the linear strength without relying solely on a visual chart, calculate r directly in a calculations cell:

Pearson's r
Correlation coefficient (r)=CORREL(A2:A101, B2:B101)
Coefficient of determination (R²)=RSQ(B2:B101, A2:A101)
How to interpret r in your IA. The value of r ranges from −1 to +1. In your write-up, state both the value and its meaning: "r = 0.91 indicates a strong positive linear correlation between temperature and sales." Then quote R² = 0.83 and interpret it as a percentage: "83% of the variation in sales is explained by temperature." Reference published critical values for your sample size n to test statistical significance - simply having r close to ±1 is not sufficient without this check.
Part 3

Advanced Regression Modelling

If your data displays non-linear properties, adjust the trendline model type in the Chart Editor. The table below summarises all available models and when to use each.

ModelEquation formTrendline settingTypical IA application
Linear y = mx + c Type → Linear Any direct proportional or steady-rate relationship
Exponential y = aebx Type → Exponential Population growth, viral spread, compound interest, radioactive decay
Quadratic y = ax² + bx + c Type → Polynomial, Degree → 2 Projectile trajectories, optimisation problems, parabolic arches
Cubic y = ax³ + bx² + cx + d Type → Polynomial, Degree → 3 Data with two turning points. For S-shaped growth a logistic model is usually better, but Sheets has no logistic trendline
Power y = axb Type → Power series Kepler's laws, braking distance, allometric biology
Logarithmic y = a·ln(x) + b Type → Logarithmic Diminishing returns, Richter/decibel scales, learning curves
Validation tip - do not choose a model by R² alone. Fit two or three trendline types to the same data and record each R², but do not simply pick the highest. A polynomial of higher degree will almost always raise R², and R² for exponential and power trendlines may be calculated on transformed data, so the values are not directly comparable. Also look at a residuals plot for any pattern, check that the model behaves sensibly in context (at x = 0, as x grows, outside your data range), and prefer the simpler model when the fit is similar. Present the comparison as a table and justify your choice; weighing strengths and weaknesses of approaches is the kind of critical reflection Criterion D rewards.

1 Extracting Exact Regression Coefficients

The trendline equation shown on a chart is rounded. For your IA write-up, extract the full-precision coefficients using formulas.

Linear model (y = mx + c)
Gradient m=SLOPE(B2:B101, A2:A101)
y-intercept c=INTERCEPT(B2:B101, A2:A101)
Full regression output - LINEST (array formula)
In an empty cell, type (Google Sheets spills the results into the cells around it):=LINEST(B2:B101, A2:A101, TRUE, TRUE)
Output: row 1slope (col 1), intercept (col 2)
Output: row 2standard error of the slope (col 1), standard error of the intercept (col 2)
Output: row 3R² (col 1), standard error of the y estimate (col 2)
Exponential model - linearise first (y = aebx)
Column C: take ln of y=LN(B2) - fill down to C101
Coefficient b (exponent)=SLOPE(C2:C101, A2:A101)
Coefficient a (initial value)=EXP(INTERCEPT(C2:C101, A2:A101))
Quadratic model - add x² column (y = ax² + bx + c)
Layout: x in column A, x² in column B=A2^2 - fill down to B101, and put y in column C
Run LINEST over both predictors (the x columns must be side by side and must not include y):=LINEST(C2:C101, A2:B101, TRUE, TRUE)
Row 1 gives the coefficients in reverse column order:a (the x² term, col 1), b (the x term, col 2), c (the intercept, col 3). Row 3, col 1 is R².

2 Generating Predicted Values and Residuals

Once you have your model coefficients in named cells, create a predicted values column alongside your raw data. This lets you plot the fitted curve and calculate residuals.

Example layout - linear model, m in E1, c in E2
AB (observed y)C (predicted ŷ)D (residual)
x valuesraw data=$E$1*A2+$E$2=B2-C2
↓ fill down↓ fill down↓ fill down↓ fill down

Plot columns A and C together (alongside A and B) on the same scatter chart to show how well your model fits the data. A residuals plot (A vs D) should show no clear pattern - if it does, your model is inappropriate.

Part 4

Statistical Hypothesis Tests

Before running any test, always write H₀ and H₁ explicitly in your IA. State your significance level (α = 0.05 is standard). After obtaining the p-value, write a formal conclusion.

1 Chi-Squared Test of Independence

Use when testing whether two categorical variables are independent (e.g. age group vs. preferred platform).

1
Build your observed frequency contingency table

Enter row categories in column A, column categories in row 1. Fill in observed counts. Add row and column totals using =SUM().

2
Calculate expected frequencies

For each cell: = (row total × column total) / grand total. Use absolute cell references for totals so you can fill the formula across all cells, e.g. =($G2*B$7)/$G$7.

3
Get the p-value

=CHITEST(observed_range, expected_range) - returns the p-value directly. If p < 0.05, reject H₀ and conclude the variables are not independent.

Check expected frequencies. The chi-squared test is only valid if all expected frequencies are ≥ 5. If some cells fall below this threshold, combine adjacent categories or acknowledge this as a limitation in your IA.

2 t-Test for Comparing Means

t-test formulas
Two-sample, two-tailed, equal variances assumed (use type 3 for unequal variances)=T.TEST(A2:A30, B2:B30, 2, 2)
One-sample (vs known mean μ)Calculate t manually: =(AVERAGE(A2:A101)-μ)/(STDEV.S(A2:A101)/SQRT(COUNT(A2:A101)))
Then get p-value=T.DIST.2T(ABS(t_cell), COUNT(A2:A101)-1)

3 Normal Distribution Calculations

Normal distribution formulas
Cumulative P(X ≤ x)=NORM.DIST(x, mean, stdev, TRUE)
P(a ≤ X ≤ b)=NORM.DIST(b,mean,stdev,TRUE) - NORM.DIST(a,mean,stdev,TRUE)
Inverse - find x for given probability=NORM.INV(probability, mean, stdev)
z-score for each data point=(A2-AVERAGE(A$2:A$101))/STDEV.S(A$2:A$101) - fill down
Checking normality. Calculate a z-score column as above, then create a histogram of z-scores. If the distribution is approximately normal, it should be bell-shaped and centred near zero. In your IA, describe what you observe: "The z-score histogram is approximately symmetric and bell-shaped, consistent with the assumption of normality."

4 Poisson & Binomial Distributions

Discrete distributions
Poisson P(X = k), rate λ=POISSON.DIST(k, λ, FALSE)
Poisson P(X ≤ k)=POISSON.DIST(k, λ, TRUE)
Binomial P(X = k), n trials, prob p=BINOM.DIST(k, n, p, FALSE)
Binomial P(X ≤ k)=BINOM.DIST(k, n, p, TRUE)
Part 5

Presentation Standards for Criterion B

Mathematical communication (Criterion B, 4 marks) is directly affected by how you present Sheets output in your IA document.

Show your formulas

Don't just paste a number. Write the mathematics in standard notation, say the value was calculated with Google Sheets, and give the result: "The correlation coefficient, calculated using Google Sheets, is r = 0.87." A screenshot of your spreadsheet showing the formula is acceptable evidence of the method (it is software generated), but spreadsheet syntax typed into your own text is not standard mathematical notation.

Consistent decimal places

Decide on a precision and stick to it. For example, use 3 or 4 d.p. for r and R². Use Format → Number → Custom to enforce this in Sheets before screenshotting tables.

Label every table

Every table needs a title above it (Table 1: Summary statistics) and column headers with units. Use Format → Alternating colors for clean Sheets tables before exporting.

Clean chart styling

White background, no 3D effects, meaningful axis labels with units, a descriptive title. A pie chart rarely adds any mathematics; a bar chart usually makes comparisons easier to read.

Figure captions

Every graph in your IA needs a caption below it: Figure 2: Scatter plot of stopping distance vs. speed with quadratic regression curve (R² = 0.97). Reference the figure number in your prose.

Share your spreadsheet

Go to File → Share → Anyone with the link → Viewer if you want to share it. The IA is marked from the document you submit, so everything that earns marks (a summary of your data, the key calculations and charts) must be in the document itself, not only behind a link.

Ready to start your IA?

Use the other tools in the guide to pick a topic, understand the criteria, and avoid common mistakes.

Related IA guides

Build the maths behind your IA with practice: AA SL topics, AA HL topics, AI SL topics, AI HL topics. Looking for an idea? See IA topic ideas.