Data from two variables (x, y) can be plotted on x-y graphs as scatter plots. There are several possible patterns that can show qualitatively if there is a correlation between the two variables. A measure of the correlation is the correlation coefficient (r). To find the correlation coefficient, place the x-values in L1 and the y-values in L2. Then STAT TESTS LinRegTTest, being sure that the Lists for x an y are correct. Scroll down to find “r”. The values of “r” are 1 r 1 . A negative value indicates a negative correlation. (as x increases y increases) A positive value indicates a positive correlation (as x increases y increases) To find if there is a Correlation to some significance (probability of being wrong), we need to compare the Test Statistic “r” to a Critical Value. To find the Critical Value, need to use “n” (the number of ordered pairs (x, y)) and a table for the given significance. Table II is such a table for 0.05 . There are other tables for other significance, but we will only use 0.05 and only Table II. If |r|≥ Critical Value, then there is correlation. If there is a correlation, then there is a line that represents the relationship. The line can be approximated from the sample of (x, y) pairs. A linear relationship has the form y=mx+b. This line is only an approximation of the regression line and will be given by y ax b where a is the approximate slope and b is the approximate y-intercept. To find the regression line, use the same function as for finding “r” and find the “a” & “b”. If “r” is positive then “a” will be positive. If “r” is negative then “a” will be negative. The equation y ax b will give an approximate value of y for any given x, but because it is approximate it will have an associated error. The error can be found for any value of x that is in the (x, y) data. The As error is called the residual and is y y . an example if one of the (x, y) pairs is (2, 5) and y 2 x 3 then for x = 2 y 7 . Then the residual is 5 – 7 = -2. Two Cautions: 1. Correlation does NOT imply Causality: Just because one variable goes up as another goes up, does not mean that one causes the other to go up. 2. The scope of the Regression Equation is only the interval of the data from the ordered pairs (i.e. the domain is from x-min to x-max. Outside of this interval we can not determine what the relationship is. Do first problem of Lab 10.