Skip to content
Omnikalc

BIC model comparison calculator

Enter the number of observations and the residual sum of squares and the number of parameters of two regression models to get their BIC and AIC, the difference and the approximate probability of each model.

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

Difference BIC(B) − BIC(A)

5.48

Positive evidence for model A
BIC of model A
105.44
BIC of model B
110.92
Approximate probability of model A
93.9 %
AIC of model A
97.63
AIC of model B
95.29

The BIC model comparison calculator compares two regression models with the Bayesian information criterion, and gives the AIC, the difference, the evidence strength and the approximate probability of each model.

How the calculation works

  • Enter the number of observations.
  • Enter the residual sum of squares and the number of parameters of each model.
  • The calculator gives the BIC and AIC of both models, the difference, and the probability of model A.

Example

With 100 observations, model A has RSS 250 and 3 parameters, and model B has RSS 230 and 6 parameters. The BIC is 105.44 for A and 110.92 for B, a difference of 5.48, so the BIC prefers A with a probability of about 93.9 %. The AIC is 97.63 against 95.29, and prefers B: the criteria disagree because the BIC penalizes complexity more.

Why it matters

The calculator helps students, researchers and analysts who choose between nested or competing regressions and want a quick comparison from the output of their software.

Good to know

The formulas assume models fitted by least squares with normally distributed errors and compared on the same data. The number of parameters should include the intercept, and some software also counts the error variance, which changes both models by the same amount, so the difference is unaffected. The BIC is an approximation to the log of a Bayes factor, valid for large samples, and the probabilities are rough. Do not compare models fitted on different responses or with a different number of observations. The criteria are not tests: a low BIC says a model is best among those tried, not that it is correct. Check the residuals and the use case too.

The linear regression calculator fits a line, the R-squared calculator measures fit, the Bayes factor calculator compares hypotheses and the statistical power calculator plans sample sizes.

Frequently asked questions

What is the BIC?

The Bayesian information criterion is n ln(RSS / n) + k ln n for a regression with normal errors. The first term rewards fit and the second penalizes each parameter by ln n. The model with the lowest BIC is preferred.

How strong is a difference?

A common scale says that a difference between 0 and 2 is weak, 2 to 6 positive, 6 to 10 strong and above 10 very strong evidence for the model with the lower BIC. The probability shown assumes equal prior odds for the two models.

Why do the BIC and the AIC disagree?

The AIC penalizes each parameter by 2, the BIC by ln n, which is larger as soon as n exceeds 7. So the BIC favors simpler models, especially for big samples, while the AIC accepts more complexity to improve prediction.

Was this calculator helpful?

Thanks for your feedback!

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