Skip to content
Omnikalc

DFFITS and studentized residual calculator

Enter the number of observations, the number of parameters, the leverage, the raw residual and the residual mean square of a regression to get the studentized residual, DFFITS and the thresholds that flag an outlier or an influential point.

Updated on
Your inputs

Value out of range

Value out of range

Value out of range

Value out of range

Value out of range

Results

Externally studentized residual

2.5298

Possible outlier at the 5 % level

DFFITS

1.4606

Influential: removing this point changes its own fitted value a lot
Internally studentized residual
2.3094
Critical value at 5 %
2.056
DFFITS threshold 2·√(p/n)
0.6325
Leverage threshold 2p/n
0.2000
High leverage: the point is far from the others in the predictors
Two-sided p-value of the residual
0.0178

The DFFITS and studentized residual calculator tells you whether a point in a regression is an outlier and whether it is influential, from its leverage, its residual and the residual mean square.

How the calculation works

  • Enter the number of observations and parameters.
  • Enter the leverage, the raw residual and the MSE.
  • The calculator gives the studentized residuals, DFFITS, the thresholds and the p-value.

Example

A model with 3 parameters is fitted to 30 points. One point has a leverage of 0.25, a residual of 2.4 and the MSE is 1.44, so s = 1.2. The internal residual is 2.4 / (1.2 × √0.75) = 2.3094, the external one is 2.5298 against a 5 % threshold of 2.056, and DFFITS is 1.4606 against 0.6325. The point is both an outlier and influential.

Why it matters

The calculator helps students and analysts who check regression assumptions, data scientists who clean data sets, and researchers who must justify keeping or removing a point.

Good to know

Never delete a point only because a diagnostic flags it. Check for a data entry error, a different population or a model that misses a variable first, and report results with and without the point. Cook’s distance combines the same ingredients, and the two diagnostics often agree. The thresholds are rules of thumb, and with many points, a few will exceed them by chance. With large samples, use a Bonferroni correction on the residual test. Leverage depends only on the predictors, not on the response, so a point can have a high leverage and a small residual. Such a point is not harmful unless it is also off the trend.

The Cook’s distance calculator gives another influence measure, the linear regression calculator fits a line, the R-squared calculator measures fit and the t-distribution calculator gives critical values.

Frequently asked questions

What is a studentized residual?

It is a residual divided by an estimate of its own standard deviation, which makes residuals comparable. The externally studentized version estimates the error without the point in question, so a real outlier does not hide itself by inflating the error.

What does DFFITS measure?

It is the change in the fitted value of an observation, in standard errors, when that observation is left out of the fit. A common rule flags |DFFITS| above 2 × √(p / n).

Where do I get the leverage and the MSE?

Statistics software gives the leverage (hat value) of each point and the residual mean square of the model. The leverage lies between 1/n and 1, and the leverages add up to the number of parameters.

Was this calculator helpful?

Thanks for your feedback!

Results are estimates provided for information purposes. They do not replace professional advice.