Log Mode

Enabling log mode changes the strategy that the calculator uses to fit regression parameters. By default, regression parameters are chosen to minimize the sum of the squares of the differences between the data and the model predictions. When log mode is enabled, a transformation that makes the model linear is applied to both the data and the model before fitting the parameters.

Log mode can be enabled for models of the following forms:

  • Exponential: y_1 \sim ab^{x_1}, \space y_1 \sim a\exp(bx_1), \space y_1 \sim 2^{ax_1+b}
  • Power Law: y_1 \sim ax_1^b
  • Logarithmic: y_1 \sim a\ln(x)+b

When log mode is enabled, the r and r^2 statistics are calculated according to the transformed data, but the residuals are calculated according to the untransformed data.

Compatibility with other tools

Several other math software systems use procedures similar to log mode by default. For example, TI calculators use logarithmic transformations for ExpReg, PwrReg, and LnReg, Geogebra uses logarithmic transformations for FitExp, FitGrowth, FitLog, and FitPow, and Excel uses logarithmic transformations in its Exponential, and Power regression lines. Enabling log mode in Desmos makes it easier to compare answers with these other systems.

Why doesn't Desmos use log mode by default?

Many math software systems only allow regressing a handful of models and have different function names for each model, for example, ExpReg, PwrReg, LnReg, etc.

Desmos is significantly more flexible: any relation between lists of data can be used as a regression model. For example, after regressing a model of the form

y_1 \sim ax_1^b

you can immediately check the effect of adding a constant by changing the model to

y_1 \sim ax_1^b+c

The first model can be linearized by a logarithmic transformation, but the second model cannot. Desmos defaults to using least squares regression without transforming the data so that the procedure used to fit these two models is the same.

Have more questions? Submit a request

4 Comments

  • 6
    Avatar
    Owen J

    I think that it is very cool

  • 0
    Avatar
    Mike Rodgers

    I was looking for a way to plot a semi-log graph.   That is one axis scale is logarithmic.   Handy for functions that growl exponentially.

    Edited by Mike Rodgers
  • 0
    Avatar
    Adrian Nguyen

    How do you even enable log mode?

  • 0
    Avatar
    John Bennett

    When you enter the model, for example, y1 ~ C * a^(x1), and a check box will appear for enabling log mode.

Please sign in to leave a comment.
Powered by Zendesk