Minitab-相关性与回归分析.ppt

修改履历 7 Explain difference/similarity between r and R-squared. R-squared is a measure of the tightness of fit. Linear Regression:Detailed measure of association between an independent (X) and a dependent (Y) variable. Output of a linear regression is an equation of a line that represents the linear relationship between X and Y. Correlation: Measure of strength of association between two quantitative variables that are in a logical numeral sequence (can be KPIVs or KPOVs). R2 or r Coefficients: Measures the proportion of variability that two variables share or the strength of the linear relationship. R2 is in the range 0 to 1 and r is in the range of -1 to +1. Begin by defining correlation, and use it to bridge to regression as a more detailed version of correlation. Emphasize the independent and dependent variables are only used in regression. State that correlation and regression are a first hand look at data and may not account for mathematical variations such as 值s related by logs, square, cubes, etc. 通常是看P值啦! 8 Define what a good and bad R-squared 值 should be. Recommend use of this chart to show others what the 值 means in the future. Emphasize that good judgment must prevail in interpretation because of the “short term” look taken in correlation and regression. R2 = 1 - SSE/SST SSE = Residual point - pt on the line SST = Total error, total variability in the whole data set 22 ## JRA Correlation 8/97 REV 4.0 2/9/98 Tab 25 Quiz 1) Which of the following sets of data can we try to use the tool correlation regression? Customer type Vs. Sales Volume Customer size Vs. Margin Month Vs. Inventory 2) Interpret the following: a) Is there a correlation? b) Assuming there is a correlation, what direction is it? c) If the new housing investment is estimated at $4.7MM next year, what should our sales op plan be? 3) T/F If I establish a correlation between a KPIV and KPOV I can assume that the KPIV is causes the chang