Home >> Science & Nature >> Science

How Do I Read Residual Plots in Excel?

Regression analysis is used to predict scores on an independent variable, known as "x" by using known scores on one or more dependent variables known as "y's." Linear regression analysis mathematically computes the equation for a straight line, which serves as a predictive model. According to the website, Stat Trek, residuals represent the vertical distance between each obtained data point of the independent variable from this straight line. Microsoft Excel 2007 produces a plot of the residuals which must be interpreted to assess the appropriateness of using a linear regression model.

Things You'll Need

Microsoft Excel 2007

Instructions

- 1
  Identify the x and y variables in your regression. The x variable or independent variable represents the outcome you want to measure. The y variables or dependent variables are the inputs or predictors. For example, if you want to design a model predicting number of E.R. admissions a person would have by using number of pounds overweight and number of hours worked per week, the dependent variables are number of pounds overweight and number of hours worked per week, while the independent variable is number of E.R. admissions.
- 2
  Understand that the x-axis of a residual plot contains all values of the x variable in the sample. In this example, if the highest number of E.R. admissions anyone in the sample had was 15 and the lowest was zero, the scale would start at zero and extend upward in increments of one to the maximum value of 15.
- 3
  Learn to read the y-axis of the residual plot. The y-axis represents the residuals. If the largest distance between an obtained data point and the predictive straight line is 15 and the smallest distance was zero, this scale would start at zero and extend upward in increments of one to the maximum value of 15. Microsoft Excel 2007 produces one graph for each y-variable.
- 4
  Understand that the straight line on the graph is the predictive line that describes the best-fit relationship between x and the y-variable being graphed. The line can be horizontal, slanted upward, or slanted downward depending on the nature of the relationship between x and the y being graphed.
- 5
  Look at the spread of dots above and below the straight predictive line. If there are an equal number of dots above the line as below it, linear regression is appropriate to describe the relationship between x and the y being graphed.
- 6
  Look for patterns of dispersement. If data is in clusters, a shape other than a straight line, such as a "U," or if data points are not evenly dispersed above and below the straight predictive line, linear regression is not appropriate and non-linear models must be used.

Science