linear modeling- trendlines- de paul university（线性建模-趋势线德保罗大学）.docVIP

Linear Modeling-Trendlines The Problem - To date, we have studied linear equations (models) where the data is perfectly linear. By using the slope-intercept formula, we derived linear equation/models. In the “real world” most data is not perfectly linear. How do we handle this type of data? The Solution - We use trendlines (also known as line of best fit and least squares line). Why - If we find a trendline that is a good fit, we can use the equation to make predictions. Generally we predict into the future (and occasionally into the past) which is called extrapolation. Constructing points between existing points is referred to as interpolation. Is the trendline a good fit for the data? To answer this question, you need to address the following five guidelines: Guideline 1: Do you have at least 7 data points? For the datasets that we use in this class, you should use at least 7 of the most recent data points available. If there are more data points, you will also want to include them (unless your data fails one of the guidelines below). What if you have 5 or 6 data points? It is a judgment call… we will discuss this in class. Guideline 2: Does the R-squared value indicate a relationship? R2 is a standard measure of how well the line fits the data. It is more useful in the negative sense: if R2 is very low, it tells us the model is not very good and probably shouldnt be used. The R-squared value is also called the Coefficient of Determination and can be written as r 2?or?R2. If the R2 = 1, then there is a perfect match between the line and the data points. If the R2 = 0, then there is no relationship between n the x and y values. If the R2 value is between .7 and 1.0, there is a strong linear relationship and if the data meets all the other guidelines, you can use it to make predictions. If the R2 value is between .4 and .7, there is a moderate linear relationship and the data can most likely be used to make predictions. If the R2