Free scoping read. A written assessment of your manuscript in two business days, at no cost.

Start one
Contents

Academy · Research Data, Analytics & AI · SD from Reported Z-Score

Estimating the Standard Deviation from a Reported Z-Score and Effect Estimate

A reported z-score for a mean difference converts into a pooled standard deviation (SD) in the same way as a t-value. The conversion is valid only when the z-score derives from a genuine large-sample normal-approximation test of that mean difference, not from a non-parametric statistic that is also labelled ‘z’.

Pubrica AcademyWritten by a biostatistics reviewer Reviewed bya meta-analysis methodologist Published10 October 2026 Last checked10 October 2026 Reading time7 minutes Sources10 peer-reviewed

Some papers report a z-score rather than a t-statistic for a two-group comparison, typically when the original analysis used a large-sample normal approximation instead of Student's t-distribution, or when the z-score comes from a non-parametric test such as the Mann-Whitney U test reported in its normal-approximation form. The arithmetic for converting a genuine mean-difference z-score into a pooled SD mirrors the t-value method closely, but the label "z-score" covers more than one underlying statistic, and confusing those cases is the main risk specific to this conversion.

Learning objectives

  • The direct formula for converting a mean-difference z-score into a pooled SD
  • Why a z-score reported for a non-parametric test should not be converted the same way
  • How to determine, from the information a paper reports, which kind of z-score has been reported

Common misconceptions

  • “Any statistic labelled 'z-score' in a results table can be converted into an SD the same way.” A z-score from a large-sample normal-approximation test of a mean difference estimates a different quantity than a z-score from the normal approximation to a Mann-Whitney U statistic, which reflects a rank-based comparison rather than a difference in means and SDs; converting the latter using a mean-difference formula produces a value that does not represent the quantity it purports to describe.

Conversion formula for a mean-difference z-score

When a z-score corresponds to a large-sample normal-approximation test of a mean difference between two groups, it is defined in the same way as a t-statistic, as the mean difference divided by its standard error, and the resulting formula is structurally identical1:

z = MD / SE
SE = MD / z
SDpooled = SE / √(1/n₁ + 1/n₂)

The practical difference from the t-value method is that no degrees-of-freedom adjustment applies, since a z-score by definition refers to the standard normal distribution rather than Student's t with a specific df2. This makes the z-score conversion marginally simpler mechanically, but it also means that no built-in degrees-of-freedom consistency check is available, as there is for a t-value, which places a greater verification burden on confirming which test produced the reported z.

Distinction between mean-difference and rank-based z-scores

The same label, "z," appears in biomedical papers attached to genuinely different statistics. A z-test of two means, appropriate for large samples where the normal approximation to the sampling distribution of the mean difference is reasonable, produces a z-score that estimates exactly the quantity this conversion formula assumes1. A Mann-Whitney U test, by contrast, is a non-parametric rank-based comparison, and when a paper reports its result using the large-sample normal approximation, the resulting "z" reflects a standardised rank-sum statistic, not a standardised mean difference, even though it is reported with the identical notation23.

A practical indicator is the set of statistics the paper reports alongside the z-score: a mean-difference z-test is usually reported with means (or a mean difference) for each group, whereas a Mann-Whitney result is more often reported with medians, ranks, or an explicit mention of the Mann-Whitney or Wilcoxon rank-sum test by name. Checking the test named in the paper's methods section, not only the symbol reported in the results table, is the most reliable way to resolve this before any conversion is attempted2.

Implications of the distinction for pooled analysis

Applying the mean-difference conversion formula to a Mann-Whitney-derived z-score produces a value with the units and general form of an SD, but it does not estimate the spread of the underlying continuous outcome; instead, it estimates a quantity related to the variance of a rank-sum statistic, which the Cochrane Handbook treats as conceptually distinct from a standard mean-difference SD42. Entering that mismatched figure into a standard mean-difference meta-analysis alongside directly reported SDs from other studies would combine two statistically different quantities under the same label, which can distort a pooled estimate in a manner that is difficult to detect retrospectively, since the resulting value appears superficially plausible.

Errors of this kind, which are difficult to detect, illustrate the value of a documented extraction process: recording, for every included study, which specific test produced each reported statistic, and not only its numeric value, gives a later reviewer or an editor the information needed to identify such a mismatch before it reaches a published pooled estimate5.

Correct application of the conversion in data extraction

In practice, before converting any reported z-score, a reviewer should locate the description in the methods section of the test used for the specific comparison in question, confirm that it was a comparison of means (not medians or ranks), and only then apply the mean-difference conversion formula12. Where the test is confirmed as a large-sample z-test of means, the conversion proceeds as shown above, with the resulting SD labelled as derived rather than directly reported.

Where the underlying test cannot be confirmed, or is confirmed to be a rank-based test, the study should either be excluded from a mean-difference meta-analysis on this outcome or handled through a synthesis method designed for mixed or non-parametric data, an approach consistent with PRISMA 2020's broader expectation that a review's synthesis methods be matched to what the data of the included studies can support rather than forced into a single framework regardless of fit67.

Worked example of correct and incorrect application

Consider a large trial of n₁=n₂=150 that reports a mean difference of 3.1 units in a continuous biomarker, with the methods section explicitly stating that a two-sample z-test was used given the large sample, and a reported z-score of 2.44. Once it is confirmed that the methods section names a z-test of means rather than a rank-based procedure, the conversion proceeds directly: SE = 3.1/2.44 ≈ 1.27, and SDpooled = 1.27 / √(1/150+1/150) = 1.27/√0.0133 = 1.27/0.1155 ≈ 11.01.

By contrast, consider a different paper that compares the same two sample sizes on an ordinal pain-score outcome and reports “Mann-Whitney U test, z=2.44” with group medians rather than means. Although the reported z-score is numerically identical to that in the first example, applying the same formula here would be a direct misuse: this z reflects a standardised rank-sum statistic from a non-parametric comparison of distributions, not a standardised mean difference, and the paper provides no mean difference for use in the formula23. The only appropriate course for this second study is to exclude it from a mean-difference pooled analysis on this outcome or to convert its reported medians and ranks using a method designed for that purpose; its z-score should never be forced through the mean-difference formula merely because a value is available.

Relationship to the t-value and p-value conversions

A mean-difference z-score occupies a conceptual position between the t-value and p-value conversion methods covered elsewhere in this cluster. Like a t-value, a z-score is a direct ratio of the mean difference to its standard error, so recovering the SE requires only the single-line rearrangement SE=MD/z, with no inverse-distribution calculation; the p-value method, by contrast, must first pass through the inverse of the normal distribution function to obtain an implied z-score before the same rearrangement can be applied14.

Unlike a t-value, however, a z-score carries no accompanying degrees-of-freedom figure to serve as a built-in consistency check against the reported sample sizes, since the standard normal distribution has no degrees-of-freedom parameter1. A reviewer extracting from a z-score should therefore rely more heavily on the qualitative check of confirming the underlying test by name in the methods section, since the quantitative degrees-of-freedom check available for a t-value is not available here.

One substitute check is plausibility against the reported sample sizes: a z-test is generally considered reasonable only when both groups are reasonably large, with a rough threshold of at least 30 observations per group often cited, so a reported z-score for a comparison with notably smaller samples warrants investigation, either because the normal approximation was a weaker choice in the original analysis or because the reported statistic may not be what it appears to be14. This threshold is a rule of thumb rather than a strict cut-off, but it is a useful initial filter when deciding how much scrutiny a reported z-score requires before conversion. Combining this plausibility check with the methods-section confirmation described earlier gives a reviewer two independent, complementary means of identifying a mismatched statistic before it reaches a pooled analysis.

Key points

Key points

A z-score from a genuine large-sample mean-difference test converts into a pooled SD in the same way as a t-value (SE=MD/z), whereas a z-score from a non-parametric test such as the Mann-Whitney test estimates an entirely different quantity and should not be converted using this formula.

Practical application

Practical application

  1. The methods section of the paper should be checked for the name of the test, not only the reported z symbol. A z-test of means and a normal-approximation Mann-Whitney result are both labelled ‘z’ but estimate different quantities.
  2. The comparison should be confirmed as one of means, not medians or ranks, before conversion. The reporting of group means (or a mean difference) rather than medians is a practical indicator.
  3. SE = MD/z and SD_pooled = SE/sqrt(1/n1+1/n2) should be applied only once the test type is confirmed. No degrees-of-freedom adjustment is needed, unlike the t-value method.
  4. Any z-score that cannot be confirmed as deriving from a mean-difference test should be excluded or handled separately. Entering an unconfirmed or rank-based z-score into this formula risks an undetected error in a pooled analysis.

Frequently asked questions

Mechanically, the formulas are structurally identical (SE=MD/statistic, then SD_pooled from SE), but a z-score requires no degrees-of-freedom check since it refers to the standard normal distribution; the principal additional risk with a z-score lies in confirming that it derived from a mean-difference test rather than a rank-based one.

Not using this mean-difference formula; it would require a different approach suited to rank-based statistics, and in most cases the more defensible choice is to treat the study's outcome through a synthesis method designed for non-parametric or ordinal data rather than forcing it into a mean-difference framework.

Secondary indicators should be sought: accompanying means versus medians, mention of normality testing before a parametric test was chosen, or a stated justification for the chosen test; if the test type cannot be determined, the safer choice is to exclude that study's z-score from this conversion.

The method assumes that the normal approximation underlying the original z-test is itself reasonable, which generally holds better in larger samples; for a z-test applied to a small sample, the normal approximation may already have been a weaker choice in the original analysis, a limitation that carries through to this derived SD.

References

  1. 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
  2. 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
  3. Altman DG, Bland JM. Standard deviations and standard errors. BMJ. 2005;331(7521):903. read 10 October 2026. View on The BMJ
  4. 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
  5. 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
  6. 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
  7. Mulrow CD. Rationale for systematic reviews. BMJ. 1994;309(6954):597-599. read 10 October 2026. View on The BMJ
  8. 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
  9. 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
  10. Walter SD, Yao X. Effect sizes can be calculated for studies reporting ranges for outcome variables in systematic reviews. Journal of Clinical Epidemiology. 2007;60(8):849-852. read 10 October 2026. View on Journal of Clinical Epidemiology

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 Z-Score and Effect Estimate. Pubrica. https://pubrica.com/academy/calculation-of-standard-deviation-only-with-z-score/

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.

Z-Score-to-SD Conversion Worksheet

This checklist applies when a source reports a z-score and mean difference but no SD, SE, or CI.

[ ] Located the methods section's named test for this specific comparison
[ ] Confirmed the comparison is of means, not medians or ranks (check for Mann-Whitney/Wilcoxon naming)
[ ] Computed SE = MD/z
[ ] Computed SD_pooled = SE / sqrt(1/n1+1/n2)
[ ] Labelled the resulting SD as derived, noting the confirmed test type 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 checks the named test behind every reported z-score before converting it into a pooled SD, so that rank-based statistics are not misclassified as mean-difference SDs in a client’s meta-analysis.

Find something else

Search the Pubrica Academy

Access is free and requires no account. Enter a topic, or open the full library to filter by category, sub-topic or resource type.

Verification that a reported z-score derives from a mean-difference test

A scoping assessment is free and takes two business days. It provides a written assessment of what the work requires and which specialist would undertake it.

Get a free quote

Regulations

Check your project against this rule

  1. 1Your work
  2. 2Your details
  3. 3Send
What do you need?

Material you send is held under signed confidentiality, with access limited to the assigned team and encrypted transfer. It is not used to train anything and is not passed to third-party tools. Ask for the agreement before you send anything you would rather cover first.