• Clinicallydepressedpoochie@lemmy.worldOP
            link
            fedilink
            arrow-up
            1
            arrow-down
            3
            ·
            edit-2
            4 days ago

            The answer is you cant know that’s why it would be crazy to model it that way. With the infinite possible exponential functions that could fit this curve.

            • mxeff@feddit.org
              link
              fedilink
              arrow-up
              2
              ·
              4 days ago

              But isn’t the same true also for a linear model, which of the infinite possible linear functions could fit this curve?

              • Clinicallydepressedpoochie@lemmy.worldOP
                link
                fedilink
                arrow-up
                1
                arrow-down
                3
                ·
                4 days ago

                Nope. There isn’t an infinite number of linear functions that will fit this curve. Based on the slope and y intercept there is only one linear function. There may be other approximations but only one line of best fit that will predict future outcomes.

                • mxeff@feddit.org
                  link
                  fedilink
                  arrow-up
                  2
                  ·
                  4 days ago

                  But this takes only into consideration the two selected point used to calculate the the slope and intercept. All other point will not exactly lie on linear function. And as you can choose any combination of two point you will get again infinitely many different parametrization a of the linear model.

                  • Clinicallydepressedpoochie@lemmy.worldOP
                    link
                    fedilink
                    arrow-up
                    1
                    arrow-down
                    2
                    ·
                    edit-2
                    4 days ago

                    But if it is a linear function it doesnt matter if you only use two points. Unless the function is not linear. If the line isn’t completely smooth you could also use a polynomial function that would fit the curve. Unless it exhibits exponential behavior why would you ever model it as an exponential if linear or polynomial models the data more accurately?