Some links on this page are affiliate links: if you buy through them we may earn a commission, at no extra cost to you.
Yes. Google Sheets can handle descriptive statistics, grouped summaries, charts, correlations, basic regression, confidence intervals and t-tests. It works well for transparent, collaborative analysis of small-to-medium datasets. It is not a substitute for specialist software when you need complex models, extensive diagnostics or a reproducible research pipeline. The key is to match the method to the data and study design—not just find a formula that returns a number.
Start with a clean, auditable dataset
Statistical analysis means more than calculating an average. A sound workflow moves from preparing data to summarizing it, examining patterns, testing a suitable question and communicating uncertainty. Keep the original data intact, and put formulas and outputs on a separate analysis sheet.
![Google Workspace Guide: Unlock Every Google App – Elevate Efficiency with Exclusive Tips, Time-Savers & Step-by-Step Screenshots for Quick Mastery [II EDITION]](https://m.media-amazon.com/images/I/41CX0AjJhAL._SL160_.jpg)
Use one row per observation and one column per variable, with a clear header row. For example, a dataset might have Record ID, Group, Date, X variable and Y variable columns. Avoid merged cells within the data range. Make sure numbers are stored as numbers and dates as dates; text that looks numeric can be omitted or treated differently by formulas. Record what blanks mean, and do not replace missing values with zero unless zero is genuinely the recorded value. Check duplicates and inconsistent category labels rather than removing records automatically.
Quick wins for a faster PC:
Scan for outdated or missing drivers - takes under a minuteDriver Scan →Repair Windows errors before they cause bigger problemsFix Now →Useful preparation functions include FILTER, SORT, SORTN, UNIQUE and QUERY. For example, to extract rows in which column B says Treatment:
#1 Best Overall
- hole punched
- high quality card stock
- 4 pages
- made in USA
- keyboard shortcuts
=FILTER(A2:E, B2:B="Treatment")To summarize the average of column E by the category in column B:
=QUERY(A1:E, "select B, avg(E) where B is not null group by B label avg(E) 'Average outcome'", 1)QUERY uses Google Visualization API Query Language, with its own syntax and limitations; it is not general-purpose SQL. See Google’s Sheets guidance on functions, pivot tables and charts.
Build a descriptive-statistics summary
Assume numeric observations are in B2:B101. These formulas cover common first questions:
Recommended Free Tools
| What you want | Formula |
|---|---|
| Count of numeric observations | =COUNT(B2:B101) |
| Count of non-empty cells | =COUNTA(B2:B101) |
| Mean | =AVERAGE(B2:B101) |
| Median | =MEDIAN(B2:B101) |
| Mode | =MODE(B2:B101) |
| Minimum and maximum | =MIN(B2:B101) and =MAX(B2:B101) |
| Range | =MAX(B2:B101)-MIN(B2:B101) |
| First and third quartiles | =QUARTILE(B2:B101,1) and =QUARTILE(B2:B101,3) |
| 90th percentile | =PERCENTILE(B2:B101,0.90) |
| Sample standard deviation | =STDEV.S(B2:B101) |
| Population standard deviation | =STDEV.P(B2:B101) |
Use the mean when data is reasonably symmetric and not dominated by extreme values. The median is more resistant to skew and outliers. A mode can be useful for repeated categories or discrete measurements; it may not be informative for continuous measurements.
Choose STDEV.S and VAR.S when the observations are a sample from a larger population of interest. Choose STDEV.P and VAR.P when the data covers the complete population you intend to describe. Google documents STDEV as the sample standard deviation equivalent to STDEV.S; population alternatives include STDEVP and STDEV.P (function details). This choice is only a calculation convention: it does not make a poor sampling design valid or resolve dependence between observations.
Summarize groups with formulas or pivot tables
For a single group, criteria functions avoid manually filtering the source. If group labels are in B2:B101 and outcomes are in E2:E101:
=COUNTIF(B2:B101, "Treatment")
=AVERAGEIF(B2:B101, "Treatment", E2:E101)
=AVERAGEIFS(E2:E101, B2:B101, "Treatment", C2:C101, ">="&DATE(2026,1,1))
=MEDIAN(FILTER(E2:E101, B2:B101="Treatment"))For criteria-based functions, keep criteria ranges and measurement ranges aligned row-for-row. A median or other summary for several groups can be built from a group list, a pivot table or a QUERY summary.
Do these 3 things before closing this tab:
1Fix the driver behind crashes, sound loss and screen glitches2Clear out junk files and repair common Windows errors3Scan for outdated or missing drivers - takes under a minuteTo create a pivot table in the desktop interface, select the source data and choose Insert → Pivot table. Choose where to place it, then add fields under Rows, Columns, Values and, if useful, Filters. You can summarize values as counts, sums or averages. Google’s pivot table instructions describe this workflow; labels can vary with platform, language or interface changes.
Rank #2
- Mastering Google Sheets: A Step by Step Handbook for Beginners to Simplify Data Analysis, Boost Productivity, and Unlock Your Full Spreadsheet Potential
- ABIS BOOK
Pivot tables are useful for questions such as average sales by region, responses by category or outcomes by treatment group. They are descriptive summaries, not a complete inferential analysis: they do not automatically provide suitable confidence intervals, control for confounding variables or establish that an observed difference is statistically significant or causal.
Choose charts to inspect the data
Select the relevant range and choose Insert → Chart, then check the chart type, X-axis, series, titles, units and legend in the Chart editor. Google documents the insertion workflow and common chart types in its chart guidance.
- Bar or column chart: compare categories.
- Line chart: show a trend over time when dates are in order.
- Scatter chart: inspect the relationship between two numeric variables.
- Histogram: examine the distribution of a numeric variable.
For scatter charts, inspect the points before calculating correlation or fitting a trendline. Look for curvature, clusters, outliers, unequal spread, a restricted range or a pattern driven by one observation. A trendline can help reveal a pattern; it does not prove causation or guarantee that a relationship will continue. Google describes scatter charts and their customization here, and documents trendline and error-bar options under chart customization here.
Avoid unordered categories on a line chart, truncated bar-chart axes that exaggerate differences, dual axes that imply a relationship between unrelated scales, and percentages without denominators. Check that dates are true dates, not text. A chart is an aid to inspection and communication, not a substitute for analysis.
Measure association with correlation
For numeric variables in D2:D101 and E2:E101:
=CORREL(D2:D101, E2:E101)CORREL returns the Pearson product-moment correlation coefficient. Values nearer +1 indicate stronger positive linear association; values nearer -1 indicate stronger negative linear association; values near zero indicate little linear association. The official statistical function list also includes covariance functions and RSQ.
Pearson correlation describes linear association, not every possible relationship. A value near zero can accompany a strong curved relationship. Outliers can change the coefficient substantially, and repeated or clustered observations may violate independence. Correlation does not establish that one variable causes another. Show a scatter plot with the coefficient, and consider whether the magnitude matters in practice rather than treating the number alone as the conclusion.
Fit a basic linear regression
For outcome Y in E2:E101 and predictor X in D2:D101, Sheets offers:
=SLOPE(E2:E101, D2:D101)
=INTERCEPT(E2:E101, D2:D101)
=RSQ(E2:E101, D2:D101)
=STEYX(E2:E101, D2:D101)These return the slope, intercept, squared Pearson correlation and standard error of predicted Y values, respectively. For a prediction using a value of X in D2:
=INTERCEPT($E$2:$E$101,$D$2:$D$101)+SLOPE($E$2:$E$101,$D$2:$D$101)*D2Or use =FORECAST.LINEAR(D2,$E$2:$E$101,$D$2:$D$101). For more regression output, use:
=LINEST(E2:E101, D2:D101, TRUE, TRUE)LINEST fits a least-squares linear trend; its final TRUE requests additional statistics. The output is an array, so leave room for it and label the returned values rather than treating the block as one number. Google’s LINEST documentation describes its regression statistics.
LINEST can accept multiple predictor columns, for example =LINEST(E2:E101,D2:F101,TRUE,TRUE). The columns are predictors, and coefficient interpretation depends on their order. Multicollinearity can make estimates unstable. Before relying on a model, inspect the scatter plot and residuals; consider linearity, influential points, constant variance, independence, missingness and whether there are enough observations for the number of predictors. A high R-squared does not validate the assumptions, show that the model will predict well outside the sample, or establish a causal effect. Sheets can calculate basic regressions, but it is not a full diagnostics and reporting environment.
Compare two groups with a t-test
Google Sheets uses =T.TEST(range1,range2,tails,type). The two ranges must have the same number of data points. Set tails to 2 for a two-tailed test or 1 for a one-tailed test. Set type to 1 for paired data, 2 for two independent samples with equal variance, or 3 for two independent samples with unequal variance. See Google’s T.TEST documentation.
=T.TEST(B2:B21, C2:C21, 2, 3)This is a two-tailed, unequal-variance test for independent groups. For matched before-and-after observations, where each row is the same subject or a genuine matched pair:
=T.TEST(B2:B21, C2:C21, 2, 1)Equal range lengths do not make data paired; pairing comes from the study design. Do not choose a one-tailed test after observing the result—its direction must be specified in advance. Testing many groups or subgroups can also inflate false-positive risk if multiple comparisons are not addressed.
The result is a p-value conditional on the test assumptions. It is not the probability that the null hypothesis is true, a measure of effect size or proof that a difference matters. Report group sizes and descriptive statistics alongside the test; include the difference in means and an appropriate uncertainty interval when possible. A statistically detectable result can still be practically trivial.
The Tool Desk
Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →Show uncertainty with confidence intervals
Sheets includes CONFIDENCE.T and CONFIDENCE.NORM. For a common t-based interval around a sample mean, calculate the lower and upper bounds as:
Rank #4
![The Google Workspace Bible: [14 in 1] The Ultimate All-in-One Guide from Beginner to Advanced | Including Gmail, Drive, Docs, Sheets, and Every Other App from the Suite](https://m.media-amazon.com/images/I/41AWmhmGFuL._SL160_.jpg)
- The Google Workspace Bible: [14 in 1] The Ultimate All in One Guide from Beginner to Advanced Including Gmail, Drive, Docs, Sheets, and Every Other App from the Suite
- ABIS BOOK
=AVERAGE(B2:B101)-CONFIDENCE.T(0.05,STDEV.S(B2:B101),COUNT(B2:B101))
=AVERAGE(B2:B101)+CONFIDENCE.T(0.05,STDEV.S(B2:B101),COUNT(B2:B101))Here, alpha 0.05 corresponds to a 95% confidence level in this setup. The t-based interval requires conditions that make the method reasonable; substantial skew, dependence or an unusual sampling design may call for another approach. In frequentist terms, 95% describes the long-run coverage of the procedure, not a literal probability that this already-calculated interval contains a fixed parameter. A narrow interval can still describe an effect too small to matter.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Distributions, simulation and time series
The statistical function list includes distribution and inverse-distribution functions such as NORM.DIST, NORM.INV, T.DIST, T.INV, CHISQ.DIST, BINOM.DIST, POISSON and WEIBULL (Google’s function list). For example, =NORM.DIST(x,mean,standard_deviation,TRUE) calculates a cumulative normal probability. A simple random normal draw can be written =NORM.INV(RAND(),mean,standard_deviation).
Simulation can illustrate sampling variation or model a scenario, but RAND() recalculates, so results can change. Copy and paste values if you need a stable snapshot, and record the assumptions. A simulation is only as credible as its inputs and model.
For time-series work, sort and validate dates, decide how to group observations by week or month, and distinguish trend, seasonal patterns and noise. A moving average such as =AVERAGE(B2:B8) smooths a seven-row window; it is not a seven-day average unless the rows represent consecutive days. TREND(known_y,known_x,new_x) estimates a linear trend. Line charts are useful for time-based patterns, but nearby observations are often correlated. A simple t-test or regression that assumes independent observations may understate uncertainty, and extrapolating a trend into the future is risky.
Use Gemini as an assistant, not a statistician
Google says Gemini in Sheets can help generate formulas, analyze data, create charts and build pivot tables, subject to an eligible Google Workspace or Google AI plan; it works best with native Sheets files. Availability may depend on account and organizational settings. See Google’s Gemini in Sheets guidance.
It can suggest a formula or help explore a selected range, but verify the range, formula, sample-versus-population choice and assumptions yourself. Do not treat generated prose as a validated statistical conclusion. Check organizational privacy rules before sharing confidential or regulated data with AI features.
Common errors and how to investigate them
#DIV/0!: Check for an empty range, zero variance or a test configuration that cannot be calculated. ForT.TEST, Google notes that zero variance in both samples can produce this error.- Text-formatted numbers: Inspect cells that look numeric but are stored as text. Confirm how the chosen function handles text and blanks; do not assume they are harmless.
- Misaligned ranges: Ensure each criterion or paired value refers to the intended same row. A formula can return a plausible-looking but wrong result.
- Missing values: Determine why values are absent before excluding or filling them. Missing, not applicable and zero mean different things.
- Dates out of order: Convert text dates to actual dates and check locale assumptions before sorting or charting.
- Formula separators rejected: Locale settings can require semicolons rather than commas, and decimal conventions can differ. Adapt separators if a valid-looking formula is rejected.
- Array output collision: Functions such as
LINESTandFILTERcan return multiple cells. Clear surrounding cells and provide labeled output space. - Outliers: Investigate entry and measurement errors, but do not delete a valid extreme case just because it changes the answer. Compare results with and without influential observations when appropriate and disclose the sensitivity.
- Dynamic data: Imports, external-data functions and volatile formulas can refresh. Document retrieval dates and preserve a snapshot when the result needs to be auditable.
Wrapping a formula in IFERROR may hide an error, but it does not diagnose it. Find the cause before suppressing the display.
Outdated Drivers Are Slowing You Down
One free scan finds every outdated or missing driver and matches the right update for your exact hardware.Free scan · exact hardware matchWindows Errors? Fix Them Before They Spread
Repair common Windows errors and clear accumulated junk for a smoother, more stable PC - no reinstall needed.Free scan · no reinstallWhen Sheets is enough—and when to switch
Sheets is a practical choice for collaborative work, teaching, small-to-medium datasets, transparent calculations, descriptive summaries, simple charts, basic correlation, regression and t-tests. Pivot tables and formulas are easy to share with colleagues who already use spreadsheets.
Consider R, Python, SPSS, SAS, Stata or another specialist tool for large datasets, complex or high-dimensional models, mixed-effects or hierarchical analysis, survival analysis, generalized linear models, time-series methods that account for autocorrelation, advanced causal inference, robust standard errors, extensive diagnostics, formal reporting or a scripted, reproducible pipeline. R and Python also make it easier to preserve a repeatable sequence of data cleaning and analysis. The trade-off is a steeper learning curve and, often, a coding workflow. Excel may suit users who need a desktop spreadsheet environment or compatibility with existing files, but it is not automatically a specialist statistics package.
Sheets has statistical functions, pivot tables, charts and optional AI assistance; that is not the same as parity with a dedicated statistical environment. The formula is only one part of the analysis: design, data quality, assumptions and interpretation determine whether the answer is useful.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
Free tools Windows power users keep installed
One-click scans. No signup required.
