Written and maintained by CASRAI Editorial Board
Published · Last updated
On this page
To complete a one-way ANOVA table with missing information, use four relations that always hold in the table: the degrees of freedom add up (dftotal = n − 1 = dfbetween + dfwithin), the sums of squares add up (SStotal = SSbetween + SSwithin), every mean square is its sum of squares divided by its own df, and F is MSbetween divided by MSwithin. Start with whichever blank can be filled from a single relation, and repeat until the table is full.
Purpose of the section: what is fixed in an ANOVA table
A one-way ANOVA table is a small system of equations. Most cells are tied to other cells, so a table with several blanks can often be rebuilt exactly. The layout below follows the OpenStax Introductory Statistics 2e presentation, where k is the number of groups and n is the total number of observations across all groups.
| Source | SS | df | MS | F |
|---|---|---|---|---|
| Factor (between groups) | SSbetween | k − 1 | SSbetween / (k − 1) | MSbetween / MSwithin |
| Error (within groups) | SSwithin | n − k | SSwithin / (n − k) | (blank) |
| Total | SStotal | n − 1 | (blank) | (blank) |
The Total row has no mean square and no F. The between-groups df is the numerator df of the F statistic and the within-groups df is the denominator df. The one-way ANOVA test is always right-tailed (OpenStax, section 13.2). The NIST/SEMATECH e-Handbook of Statistical Methods states the same sum-of-squares decomposition and calls the between-group term the treatment sum of squares, with SSE = SS(Total) − SST (sections 7.4.3.1 and 7.4.3.4). For the full derivation of each sum of squares, see the CASRAI guide to the sum of squares formula.
Structure: the solving rules, in order
- Fill the Total df first if any df is given. dftotal = dfbetween + dfwithin = n − 1. Any two of these three df give the third.
- Recover n and k from the df column. n = dftotal + 1, and k = dfbetween + 1 (or k = n − dfwithin). Both routes must agree; if they do not, a df entry is wrong.
- Fill the sums of squares by addition or subtraction. SSwithin = SStotal − SSbetween, and SSbetween = SStotal − SSwithin. Any two of the three SS give the third.
- Use SS = MS × df when a row has an MS but no SS (or the reverse). MS = SS / df in each of the two non-total rows, so any two of SS, df and MS in a row give the third.
- Compute F last. F = MSbetween / MSwithin. If F is given instead, MSbetween = F × MSwithin, which unlocks the between row.
- Check every row. SSbetween + SSwithin must equal SStotal, the df must add to n − 1, and each MS times its df must return its SS.
A short annotated example
The numbers below are invented for practice, not taken from a study. A one-way ANOVA table has these entries, and everything else is blank.
| Source | SS | df | MS | F |
|---|---|---|---|---|
| Between | ? | ? | ? | ? |
| Within | ? | 20 | 8 | |
| Total | 256 | 23 |
- Total df is 23, so n = 24.
- Within row: SSwithin = MS × df = 8 × 20 = 160.
- SSbetween = 256 − 160 = 96.
- dfbetween = 23 − 20 = 3, so k = 4. Cross-check: n − dfwithin = 24 − 20 = 4, which agrees.
- MSbetween = 96 / 3 = 32.
- F = 32 / 8 = 4.00, with 3 and 20 degrees of freedom.
A second case starts from the F ratio. Suppose the table gives dfbetween = 2, dfwithin = 27, SSwithin = 270 and F = 5.0. Then dftotal = 29, n = 30 and k = 3. MSwithin = 270 / 27 = 10, so MSbetween = 5.0 × 10 = 50, SSbetween = 50 × 2 = 100 and SStotal = 100 + 270 = 370.
A third case uses only the df column. With dfbetween = 4 and dftotal = 49, there are k = 5 groups and n = 50 observations, and dfwithin = 49 − 4 = 45 (equivalently n − k = 45).
When the table cannot be completed
Some blanks are not recoverable. Knowing both df and F fixes only the ratio MSbetween / MSwithin, not either mean square, so at least one SS or MS value must be given. Likewise, two SS values with no df tell you the sizes of the sums but not k or n. If a table still has two or more unknown quantities after every rule above has been applied, the problem needs one more given cell.
Common mistakes
- Dividing SS by the wrong df. MSbetween uses k − 1 and MSwithin uses n − k. Using n − 1 for either row gives the wrong mean square.
- Computing F from sums of squares. F is a ratio of mean squares, not of sums of squares.
- Adding mean squares. Only SS and df are additive across rows. The Total row has no MS and no F, so there is nothing to check by adding MS values.
- Off-by-one in df. Total df is n − 1, not n, and between df is k − 1, not k.
- Reading F as a left-tail test. The one-way ANOVA test is right-tailed: a large F is the evidence against equal group means.
Checklist
- dfbetween + dfwithin = dftotal = n − 1.
- SSbetween + SSwithin = SStotal.
- MS = SS / df in the between and within rows, and nowhere else.
- F = MSbetween / MSwithin, with numerator df k − 1 and denominator df n − k.
- k and n recovered from two different routes agree.
To turn a completed F into a decision, compare it with a critical value or find its p-value; the guides to the F table and to interpreting the F statistic cover that step. For degrees of freedom in general, see degrees of freedom in statistics.
FAQ
Can I find F if only SS and df are missing?
Yes, as long as the mean squares are given. F = MSbetween / MSwithin needs only those two cells, so the SS and df columns are not required for the F value itself. They are required if you want to confirm the table is internally consistent.
How do I find k and n from an ANOVA table?
Read them from the df column: k = dfbetween + 1 and n = dftotal + 1. If only the within and total df are given, dfbetween = dftotal − dfwithin, and k follows from that.
How do I find the sum of squares if only the mean square and df are given?
Multiply: SS = MS × df, within the same row. For the Total row, which has no MS, use SStotal = SSbetween + SSwithin.
Does the same method work for two-way ANOVA or regression ANOVA tables?
The same ideas (SS and df add up, MS = SS / df, F is a ratio of mean squares) carry over, but the df formulas differ for each source of variation, so this page covers only the one-way layout above. For a regression ANOVA table see how to read a regression table, and for software output see one-way ANOVA in SPSS.
Sources
- OpenStax, Introductory Statistics 2e, section 13.2 “The F Distribution and the F Ratio”: https://openstax.org/books/introductory-statistics-2e/pages/13-2-the-f-distribution-and-the-f-ratio (checked 2026-10-29).
- NIST/SEMATECH e-Handbook of Statistical Methods, section 7.4.3.1 (https://www.itl.nist.gov/div898/handbook/prc/section4/prc431.htm) and section 7.4.3.4 (https://www.itl.nist.gov/div898/handbook/prc/section4/prc434.htm) (checked 2026-10-29).







