Problematic Patterns in Random NoiseProblematic Patterns in Random NoiseNo experiment is entirely without error, but when an experiment is repeated a large number of times, it is possible to approximate the true relationship between the variables with regression analysis. In essence, this analysis determines the equation of a line or curve that best approximates the data set at hand.
The first 5 Columns of this model hold the data. For each x-value in Column A, the experimenters have found a y-value in column B. Using these values, an equation is found to represent the data set. The values of this equation are shown in Column C. Finally, Column F lists the Deviation - the difference between the value of the best-fit line and the true value. The line, points, and deviations are all graphed to the right of the data.
This model is very simple. Just press [Ctrl] [=] to get a new set of random data points and a new best-fit line.
Sheet 1 contains linear data, so the line of best fit should be a good approximation for the data. Thus, your R2 value should be relatively high, and the deviations should be grouped randomly around 0.
Sheet 2 contains quadratic data, so the line of best fit should not be a good approximation for the data. Thus, your R2 value should be low, and the deviations should show a clear pattern around 0.