When I force a linear regression line to go through the origin, the 95% prediction bands seem wrong.

When you use linear regression, and check the option to force the line through a certain point (usually the origin), Prism does not create the prediction bands properly. Instead, it creates confidence bands (even if you choose prediction bands).

To work around this problem, choose the nonlinear regression analysis rather than the linear regression analysis. Then choose the "Polynomial, first order (straight line)" model, and constrain the Intercept to equal 0.0 (or some other value). Prism's nonlinear regression analysis can correctly create prediction bands for any model. Prism's linear regression analysis only creates prediction bands correctly when you don't contrain the line to go through a certain point. In contrast, the confidence bands are correct for constrained linear regression

We fixed this bug in 4.03 and 4.0c.