Free research tool

ICC calculator

Paste a table with one row per subject and one column per rater, device or session. The calculator returns all six Shrout–Fleiss intraclass correlations, ICC(1,1), ICC(2,1), ICC(3,1) and their average-measure versions, with McGraw–Wong confidence intervals, F tests and a methods sentence that names the form you chose.

Updated October 17, 2026. Free, no sign-up. Calculations run in your browser and your data is never uploaded.

1. Your data (one row per subject, one column per rater or device)

Paste data (CSV, TSV or spaces; first row may be headers)

Upload CSV

Pre-filled with the 6 targets x 4 judges example from Shrout & Fleiss (1979), Table 2. Your data stays in this browser tab; nothing is uploaded.

Columns to include

judge_1 judge_2 judge_3 judge_4

ICC form to reportPick the model that matches your design (see the guide below).Confidence level

2. Intraclass correlation

n = 6 subjects, k = 4

ICC(2,1)

0.290

Two-way random effects, absolute agreement, single measurement

95% CI

0.019 to 0.761

Koo & Li: poor to good

F test

F(5, 15) = 11.03

p = < 0.001

All six intraclass correlation forms
Shrout & Fleiss	McGraw & Wong model	ICC	95% CI	F (df)	p
ICC(1,1)	One-way random effects, absolute agreement, single measurement	0.166	−0.133 to 0.723	1.79 (5, 18)	0.165
ICC(2,1)	Two-way random effects, absolute agreement, single measurement	0.290	0.019 to 0.761	11.03 (5, 15)	< 0.001
ICC(3,1)	Two-way mixed effects, consistency, single measurement	0.715	0.342 to 0.946	11.03 (5, 15)	< 0.001
ICC(1,k)	One-way random effects, absolute agreement, mean of k measurements	0.443	−0.884 to 0.912	1.79 (5, 18)	0.165
ICC(2,k)	Two-way random effects, absolute agreement, mean of k measurements	0.620	0.071 to 0.927	11.03 (5, 15)	< 0.001
ICC(3,k)	Two-way mixed effects, consistency, mean of k measurements	0.909	0.676 to 0.986	11.03 (5, 15)	< 0.001

ANOVA mean squares: rows 11.2417, columns 32.4861, residual 1.0194, within 6.2639.

3. Methods paragraph

Reliability across 4 raters or devices (judge_1, judge_2, judge_3, judge_4) for 6 subjects was quantified with ICC(2,1), two-way random effects, absolute agreement, single measurement (Shrout & Fleiss, 1979; McGraw & Wong, 1996). ICC(2,1) was 0.290 (95% CI 0.019 to 0.761; F(5, 15) = 11.03, p < 0.001), interpreted against the bands of Koo and Li (2016) using the confidence interval. Analyses were run in the ChatPPG ICC Calculator (version 1.0; https://chatppg.com/tools/icc-calculator).

Cite this analysis

Cite the methods papers for the statistics you report, and the tool if it helped. ChatPPG. ICC Calculator (version 1.0). 2026. https://chatppg.com/tools/icc-calculator. Accessed 2026-10-18.

Preview BibTeX

@software{chatppg_icc_calculator_2026,
  author = {{ChatPPG}},
  title = {{ICC Calculator}},
  year = {2026},
  version = {1.0},
  publisher = {ChatPPG},
  url = {https://chatppg.com/tools/icc-calculator},
  urldate = {2026-10-18}
}

@article{shrout1979,
  author = {Shrout, Patrick E. and Fleiss, Joseph L.},
  title = {{Intraclass correlations: uses in assessing rater reliability}},
  journal = {Psychological Bulletin},
  year = {1979},
  volume = {86},
  number = {2},
  pages = {420--428},
  doi = {10.1037/0033-2909.86.2.420}
}

@article{mcgraw1996,
  author = {McGraw, Kenneth O. and Wong, S. P.},
  title = {{Forming inferences about some intraclass correlation coefficients}},
  journal = {Psychological Methods},
  year = {1996},
  volume = {1},
  number = {1},
  pages = {30--46},
  doi = {10.1037/1082-989X.1.1.30}
}

@article{koo2016,
  author = {Koo, Terry K. and Li, Mae Y.},
  title = {{A guideline of selecting and reporting intraclass correlation coefficients for reliability research}},
  journal = {Journal of Chiropractic Medicine},
  year = {2016},
  volume = {15},
  number = {2},
  pages = {155--163},
  doi = {10.1016/j.jcm.2016.02.012}
}

Which ICC form should I use?

The form depends on the design, so choose it before looking at the numbers (Koo & Li, 2016).

ICC(1,1) uses a one-way random-effects model: each subject is measured by a different, random set of raters. ICC(2,1) uses a two-way random-effects model for absolute agreement: the same raters or devices measure every subject, and you want results that generalise to other raters like them. ICC(3,1) uses a two-way mixed-effects model for consistency: the raters are the only ones of interest, and a constant offset between them doesn't count as disagreement.

When comparing a wearable with a reference device, systematic offsets matter, so absolute agreement, ICC(2,1), is usually the relevant form. The "k" versions apply when the score you'll use in practice is the average of k measurements.

How it is calculated

The tool runs a two-way ANOVA on the subjects × raters table to get the mean squares for subjects (MSR), raters (MSC), residual error (MSE) and within-subject variation (MSW). It then applies the Shrout and Fleiss (1979) formulas: for example, ICC(2,1) = (MSR − MSE) / (MSR + (k − 1)MSE + k(MSC − MSE)/n). Confidence intervals follow McGraw and Wong (1996), including the Satterthwaite degrees of freedom for the absolute-agreement forms.

Interpreting the value

Koo and Li (2016) suggest values below 0.5 indicate poor reliability, 0.5–0.75 moderate, 0.75–0.9 good and above 0.9 excellent, judged on the 95% confidence interval rather than the point estimate. An ICC depends on how varied the participants are: the same device can score higher in a mixed group than in a narrow one. Report it alongside Bland–Altman limits of agreement, which are in the units of measurement.

Validated against R and Python

On the Shrout and Fleiss (1979) example (6 targets, 4 judges) the calculator returns 0.17, 0.29, 0.71, 0.44, 0.62 and 0.91 for the six forms, matching the paper. Estimates, F tests and confidence intervals match R's psych::ICC and irr::icc, and pingouin in Python.

Frequently asked questions

What is the difference between ICC(2,1) and ICC(3,1)?

Both use the same two-way ANOVA. ICC(2,1) measures absolute agreement, so a constant offset between raters lowers it, and it treats raters as a random sample. ICC(3,1) measures consistency, so a constant offset doesn't count, and it treats the raters as fixed (Shrout & Fleiss, 1979; McGraw & Wong, 1996).

What is a good ICC value?

Koo and Li (2016) give below 0.5 as poor, 0.5 to 0.75 as moderate, 0.75 to 0.9 as good and above 0.9 as excellent, and recommend judging against the 95% confidence interval.

Can the ICC be negative?

Yes. A sample estimate can fall below zero when the variation within subjects is larger than the variation between them. It is usually reported as is and read as no reliability.

How should I lay out my data?

One row per subject and one column per rater, device or session, with no missing cells. Rows with a missing value are skipped and the count of skipped rows is shown.

Is my data uploaded anywhere?

No. The calculations run in your browser with JavaScript. Pasted or uploaded files are read locally and never sent to ChatPPG or anyone else, so you can use it with unpublished study data.

References

Shrout PE, Fleiss JL. Intraclass correlations: uses in assessing rater reliability. Psychological Bulletin. 1979;86(2):420-428. doi:10.1037/0033-2909.86.2.420
McGraw KO, Wong SP. Forming inferences about some intraclass correlation coefficients. Psychological Methods. 1996;1(1):30-46. doi:10.1037/1082-989X.1.1.30
Koo TK, Li MY. A guideline of selecting and reporting intraclass correlation coefficients for reliability research. Journal of Chiropractic Medicine. 2016;15(2):155-163. doi:10.1016/j.jcm.2016.02.012

More free tools

Wearable validation toolkit Bland–Altman calculator and plot Cohen's kappa calculator Bland–Altman sample size calculator SpO₂ at altitude checker All tools