Contents
On this page
- Derivation of the standard error from the t-value
- Avoidance of the normal-approximation step with a t-statistic
- Use of degrees of freedom to verify the consistency of reported values
- Application of the conversion in data extraction
- Worked example with realistic values
- Procedure when degrees of freedom are not reported
Academy · Research Data, Analytics & AI · SD from Reported T-Value
Estimating the Standard Deviation from a Reported T-Value and Effect Estimate
A reported t-statistic for a two-group mean difference can be converted directly into a standard error and a pooled standard deviation (SD). Unlike a p-value conversion, this method requires no normal-approximation step, because the t-value already incorporates the degrees of freedom used in the comparison.
Some papers report the t-statistic from an independent-samples t-test directly, alongside the mean difference between two groups, without separately reporting an SD, standard error, or confidence interval. Because a t-statistic is defined as the ratio of the mean difference to its standard error, this is arguably the most direct of the SD-estimation methods in this cluster: a simple rearrangement of that definition recovers the standard error, with no intermediate distributional conversion required.
Learning objectives
- The direct rearrangement that recovers a standard error from a reported t-statistic
- Why this method avoids the normal-approximation step required by the p-value-based method
- The degrees-of-freedom check that confirms the t-value matches the reported sample sizes
Common misconceptions
- “Conversion from a t-value and conversion from a p-value are essentially the same calculation.” A t-statistic is, by definition, the ratio of the mean difference to its standard error, so the SE can be recovered by a single-line rearrangement; a p-value, by contrast, must first be converted into an implied z-score through the inverse of the normal distribution function, an additional step with its own normal-approximation assumption that the t-value method does not require.
Derivation of the standard error from the t-value
An independent-samples t-statistic for a mean difference is defined as that mean difference divided by its own standard error1:
t = MD / SE
Because this is a direct algebraic definition rather than an approximation, rearranging it to solve for the standard error requires no additional distributional assumption beyond confirmation that a t-test was the test performed1:
SE = MD / t
Thereafter, assuming that the SDs of the two groups can reasonably be treated as pooled (a common assumption for a standard independent-samples t-test), the pooled SD follows directly from the same relationship used throughout this cluster2:
SDpooled = SE / √(1/n₁ + 1/n₂)
Avoidance of the normal-approximation step with a t-statistic
The p-value-based conversion method covered elsewhere in this cluster must pass through an inverse normal distribution calculation because a p-value alone does not specify which underlying test statistic or distribution produced it without an assumption about degrees of freedom; treating it as originating from a z-test is itself an approximation, most defensible with larger samples3. A reported t-statistic avoids this step entirely, because it is already the specific test statistic from a known distribution (Student's t, with specified degrees of freedom); no distributional conversion is therefore needed before its defining equation is rearranged14.
This makes the t-value method more accurate than the p-value method whenever both are theoretically available for the same comparison, and it is particularly valuable for smaller studies, in which the normal approximation underlying the p-value method departs most from the t-distribution used in the original analysis13. A reviewer who can extract either a reported t-value or a reported p-value for the same comparison should, for this reason, prefer the t-value.
Use of degrees of freedom to verify the consistency of reported values
An independent-samples t-test's degrees of freedom are n₁ + n₂ − 2, and many papers report this figure, explicitly or as part of a standard results notation such as "t(48) = 2.31," alongside the t-statistic itself1. Confirming that the reported degrees of freedom match the reported sample sizes provides a useful internal consistency check before the remaining extracted values are relied upon: a mismatch indicates either a typographical error in the source paper or an incorrect assignment of a reported n to the comparison, both of which should be resolved before the derived SD enters a pooled analysis4.
This check has no equivalent in the p-value-based method, since a p-value alone does not carry an explicit degrees-of-freedom figure the way a reported t-statistic typically does, which is another practical advantage of extracting from a t-value when the source paper offers both.
Application of the conversion in data extraction
In practice, a reviewer should first confirm that the reported statistic is Student's t from an independent-samples or paired comparison, rather than a superficially similar statistic such as Welch's t (which does not assume equal variances and therefore does not decompose into a simple pooled SD in the same way) or the t-value of a standardised regression coefficient25. Welch's t-test in particular is a common software default and does not assume a single pooled variance across groups, so applying the pooled-SD rearrangement to a Welch's t-value will misrepresent the quantity that the resulting number describes.
Where the degrees-of-freedom check passes and the test type is confirmed as a standard pooled-variance t-test, the resulting SD can be used with reasonable confidence, labelled as derived rather than directly reported, consistent with the same transparency expectation PRISMA 2020 sets for any statistic entering a review's synthesis through a derivation step rather than direct extraction67.
Worked example with realistic values
Suppose a paper reports, for a comparison of two independent groups of n₁=18 and n₂=20, a mean difference of 6.5 units and "t(36) = 2.85." The reported degrees of freedom, 36, match 18+20−2 = 36 exactly, which is a useful first confirmation that the statistic is indeed a standard pooled-variance independent-samples t-test and that the sample sizes have been matched correctly to this specific comparison14. The standard error follows directly: SE = 6.5/2.85 ≈ 2.28.
From there, the pooled SD is SDpooled = 2.28 / √(1/18+1/20) = 2.28 / √(0.0556+0.05) = 2.28 / √0.1056 = 2.28/0.325 ≈ 7.022. Had the reviewer instead found “t(35) = 2.85” reported for the same two sample sizes, the one-point discrepancy from the expected 36 degrees of freedom would clearly indicate that the source should be rechecked; for example, the analysed n of one group may have differed slightly from its enrolled n because of missing data, which occurs frequently enough in practice that this arithmetic check should be performed before any SD derived from a t-value is relied upon45.
Procedure when degrees of freedom are not reported
Not every paper reports degrees of freedom alongside its t-statistic, sometimes presenting only "t = 2.85, p = 0.007" without the accompanying df notation. In that situation, the reviewer can still proceed with the core conversion, SE = MD/t, since that step depends only on the t-value and mean difference, not on degrees of freedom; what is lost is specifically the consistency check, not the ability to perform the conversion itself12.
A reviewer in this position should still calculate the expected degrees of freedom from the reported sample sizes (n₁+n₂−2) and keep that figure recorded alongside the extraction, both as a reference point if the df is later needed, for instance to compute an exact p-value independently, and as a basis for judging whether the overall comparison is plausible given everything else the paper reports about its sample4. Where a paper's reported sample sizes appear inconsistent with other details in the methods section, such as a stated total enrolment that does not equal the sum of the two group sizes, that inconsistency should be resolved before any statistic derived from those numbers is relied upon, regardless of which SD-estimation method is being used.
It should also be noted that a one-sample t-test, comparing a single group's mean against a fixed reference value rather than comparing two independent groups, uses a different degrees-of-freedom formula (n−1 rather than n₁+n₂−2) and a different downstream relationship between its SE and that single group's own SD. A reviewer should confirm from the paper's methods, not only from the numeric t-value and df reported, which comparison the statistic belongs to before applying either version of this conversion, since reading the output of a one-sample t-test as though it described a two-group comparison would attribute the calculation to sample sizes that were never jointly involved in producing that value12.
Finally, where both a t-value and an exact p-value are reported for the same comparison, the two conversion methods should produce closely matching implied SDs; a meaningful disagreement between them indicates that the reviewer should recheck whether both statistics describe the same test on the same two groups before either derived figure is used in a pooled analysis16. This cross-check requires little additional effort once both figures are available, and it provides a further safeguard before a derived SD enters a pooled estimate.
Key points
Key points
A reported t-statistic converts directly into a standard error (SE=MD/t) and then into a pooled SD, with no normal-approximation step, provided that the t-value is a standard pooled-variance Student’s t, as confirmed by checking its degrees of freedom against the reported sample sizes.
Practical application
Practical application
- The reported statistic should be confirmed as a standard, pooled-variance Student’s t rather than a Welch’s t or the t of a regression coefficient. Welch's t does not assume a single pooled variance and will not decompose correctly using this method.
- Degrees of freedom (n1+n2-2) should be checked against the reported sample sizes. A mismatch indicates a data-entry or extraction error that should be resolved before proceeding.
- SE should be computed directly as MD/t, and then SD_pooled = SE / sqrt(1/n1+1/n2). No distributional conversion step is needed beyond this direct rearrangement.
- A reported t-value should be preferred over a reported p-value for the same comparison when both are available. The t-value method avoids the normal-approximation assumption the p-value method requires.
Frequently asked questions
Yes, generally. A t-statistic already encodes the specific t-distribution and degrees of freedom used in the original test, so recovering the SE from it requires no normal-approximation step, which is a source of inaccuracy in the p-value-based method, especially for smaller samples.
Welch's t-test does not assume equal variances between groups, so its t-value does not decompose into a single pooled SD using this formula; a different approach accounting for each group's separate variance would be needed instead.
A paired t-test's t-value reflects the standard error of the within-subject difference, not two independent groups' pooled SD, so the specific pooled-SD formula shown here does not apply directly; the underlying rearrangement principle (SE=MD/t) still holds, but what it estimates is the SE of the paired differences.
Because it provides a consistency check at no additional cost: if the implied degrees of freedom (n1+n2-2) do not match the reported sample sizes, something in the source or the extraction is inconsistent, and this should be identified before the derived SD is used in a pooled analysis.
References
- Higgins JPT, Li T, Deeks JJ (editors). Chapter 6: Choosing effect measures and computing estimates of effect. In: Cochrane Handbook for Systematic Reviews of Interventions, version 6.4. Cochrane, 2023. read 10 October 2026. View the Cochrane Handbook, Chapter 6
- Deeks JJ, Higgins JPT, Altman DG (editors). Chapter 10: Analysing data and undertaking meta-analyses. In: Cochrane Handbook for Systematic Reviews of Interventions, version 6.4. Cochrane, 2023. read 10 October 2026. View the Cochrane Handbook, Chapter 10
- Altman DG, Bland JM. How to obtain the confidence interval from a P value. BMJ. 2011;343:d2090. read 10 October 2026. View on The BMJ
- Altman DG, Bland JM. Standard deviations and standard errors. BMJ. 2005;331(7521):903. read 10 October 2026. View on The BMJ
- Higgins JPT, Thompson SG, Deeks JJ, Altman DG. Measuring inconsistency in meta-analyses. BMJ. 2003;327(7414):557-560. read 10 October 2026. View on The BMJ
- Page MJ, McKenzie JE, Bossuyt PM, et al. The PRISMA 2020 statement: an updated guideline for reporting systematic reviews. BMJ. 2021;372:n71. read 10 October 2026. View on The BMJ
- Hozo SP, Djulbegovic B, Hozo I. Estimating the mean and variance from the median, range, and the size of a sample. BMC Medical Research Methodology. 2005;5:13. read 10 October 2026. View on BMC Medical Research Methodology
- Wan X, Wang W, Liu J, Tong T. Estimating the sample mean and standard deviation from the sample size, median, range and/or interquartile range. BMC Medical Research Methodology. 2014;14:135. read 10 October 2026. View on BMC Medical Research Methodology
- Altman DG, Bland JM. How to obtain the confidence interval from a P value. BMJ. 2011;343:d2090. read 10 October 2026. View on The BMJ
- Mulrow CD. Rationale for systematic reviews. BMJ. 1994;309(6954):597-599. read 10 October 2026. View on The BMJ
Every source above is a peer-reviewed article, a publisher’s record of one, or a standard issued by the body that owns it. Each carries the date it was last read. Where a standard moves, this article is revised rather than left up with a note.
How to cite this article
Pick your referencing style and copy it. Your university or target journal may vary the punctuation — check its guide if it has one.
Pubrica Academy. (2026, October 10). Estimating the Standard Deviation from a Reported T-Value and Effect Estimate. Pubrica. https://pubrica.com/academy/calculation-of-standard-deviation-from-t-value/
Pubrica Academy. “Estimating the Standard Deviation from a Reported T-Value and Effect Estimate.” Pubrica, 10 Oct. 2026, pubrica.com/academy/calculation-of-standard-deviation-from-t-value/.
Pubrica Academy (2026) Estimating the Standard Deviation from a Reported T-Value and Effect Estimate. Pubrica. Available at: https://pubrica.com/academy/calculation-of-standard-deviation-from-t-value/ (Accessed: 15 October 2026).
Pubrica Academy. “Estimating the Standard Deviation from a Reported T-Value and Effect Estimate.” Pubrica. October 10, 2026. https://pubrica.com/academy/calculation-of-standard-deviation-from-t-value/
Pubrica Academy, ‘Estimating the Standard Deviation from a Reported T-Value and Effect Estimate’ (Pubrica, 10 October 2026) <https://pubrica.com/academy/calculation-of-standard-deviation-from-t-value/> accessed 15 October 2026.
Pubrica Academy. Estimating the Standard Deviation from a Reported T-Value and Effect Estimate. [Internet]. Pubrica; 2026 Oct 10 [cited 2026 Oct 15]. Available from: https://pubrica.com/academy/calculation-of-standard-deviation-from-t-value/
Pubrica Academy, “Estimating the Standard Deviation from a Reported T-Value and Effect Estimate,” Pubrica, Oct. 10, 2026. [Online]. Available: https://pubrica.com/academy/calculation-of-standard-deviation-from-t-value/
@misc{/la_standard_deviation_estimation_from_t_value_2026,
author = {{Pubrica Academy}},
title = {Estimating the Standard Deviation from a Reported T-Value and Effect Estimate},
howpublished = {Pubrica},
year = {2026},
month = {oct},
url = {https://pubrica.com/academy/calculation-of-standard-deviation-from-t-value/},
note = {Accessed: <span data-cite-acc="iso">2026-10-15</span>}
}This article is written to be learned from. Cite it if you are referring to what it says; if you are using the evidence behind it, cite the sources listed above instead.
T-Value-to-SD Conversion Worksheet
This checklist applies when a source reports a t-statistic and mean difference but no SD, SE, or CI.
[ ] Confirmed the statistic is a standard pooled-variance Student's t, not Welch's t or a regression coefficient [ ] Checked degrees of freedom (n1+n2-2) against the reported sample sizes [ ] Computed SE = MD/t [ ] Computed SD_pooled = SE / sqrt(1/n1+1/n2) [ ] Labelled the resulting SD as derived in the extraction table
Pubrica AcademyWritten and maintained by Pubrica’s subject matter experts. Articles are revised when the underlying standards change, and each article states the date on which it was last reviewed.
How Pubrica can help
Pubrica’s biostatistics review team verifies the t-test type and the consistency of the degrees of freedom before converting a reported t-value into a pooled SD, so that derived figures entering a client’s meta-analysis rest on confirmed rather than assumed test specifications.