om EJ Regression Analysis 3.1. REGRESSION Regression was introduced by backward or return to the mean value Definition. Regression is the measure of average relationship between two or more variables in terms of the original units of the data. Sir Francis Galton, Regression means moving In regression analysis there are two types of variables. The variable which is influenced or is to be predicted is called dependent variable and the variables hick influence the value or is used for prediction, is called independent variable, Independent variable is called regressor or predictor and dependent variable is called regressed ot predicted variable. Simple Linear Regression If two random variables X and Y are linearly related and mathematical functional relationship is straight line, then this linear relationship between the random variables is called simple linear regression. The straight line is called line of. regression. 3.2. LINES OF REGRESSION . There are two lines of regression for two random variables X and Y. If X is independent and Y is dependent, then line of regression is regression line of Y on X. This is used to predict the value of Y for a given value of X. In regression line of X on Y, Y is independent and X is dependent random variables and it is used to estimate the value of X for a given value of Y. Derivation of Regression Lines (® Regression line of Y on X Let us consider a bivariate distribution (x, y;), i = 1, 2, variable and x is independent variable. Let the regression line of Y on X be yratbx +), Where y is dependent_B.[Link]. Statistics (Papery 3 nino d regression coefficient of ¥ on x, 1. The slope of the regression line of Y on X is calle It is denoted by byx. Cov(XY)_ oy ea a The regression coefficient byx represents the increment in the value of dependent variable Y for a unit change in the value of independent variable X. 2. The slope of the regression line of X on Y is called regression coefficient of X on Y, It is denoted by byy. Cov (X,Y) __ ox Sr a oY ‘The regression coefficient bxy represents the increment in the value of dependent variable X for a unit change in the value of ‘independent variable Y. Properties of Regression Coefficients 1. The geometric mean between regression coefficients is the correlation coefficient Proof : We know byx ; byy= 7 SE byx byy = 7? b 1 =2VOix Oxy - 2. Ifone regression coefficient is greater than unity, then the other must be less than unity. Proof : Let us suppose that byx is greater than unity 1 ie. byx>1 > p-<1 ie. yx >1 => bs polly We know that Psi = byxbyy <1 1 ig by <5 (2) From equations (1) and (2), 1 bxy 25 5<1 bxy <1 3. Regression coefficients are independent of change of origin but not of scale. Proof : Let ap k Y =b+kV Oy = koy and ryy = ryy + byy (r do not change)Regression Analysis isis ae had 4. Arithmetic mean of the regression coefficients i ‘ Giensient ereviaea es icients is greater than the correlation Proof : We have to prove ous oxy, Consider YX + Oxy, 7 2 = 1] oy, ox 3 [rst+rs]=r = Lots ok] 2 | oxy = 08 + 0% 2 2ox0y = 0% + of ~2ox0y20 = (ox- oy? = 0 which is always true. Arithmetic mean of regression coefficients is greater than correlation coefficient. 5. Both regression coefficients must have the same sign. .e. if byx is positive, then dxy will also be positive. if by is positive, then byx will also be positive, The reason for this is sign of both regression coefficients depend on sign of the correlation coefficient only. Correlation coefficient and regression coefficients have the same sign.Since r2< 1 025 Case 1. Ifr = 0, then tan 0 = i.e. two regression lines become perpendiculay or x. Case 2. Ifr = +1, then tan0=0= 0 ie. Two regression lines regression lines passes through the point parallel. Hence if r = +1 i.e. correlation is per! regression lines coincide. 3.5. CORRELATION VS REGRESSION ‘The given below are some of basic differences either coincide or_paralle if the variables are uncorrelated. 1 to each other. But both the J), therefore regression lines are not fect positive or perfect negative, then two between correlation and regression. Correlation w] |, Correlation means the relationship between two or more variables. It measures effect of one variable due the change in the other variables. 2. Correlation need not imply cause and effect relationship between the variables under study. 3. Correlation analysis measures linear relationship only between the variables. Hence practical applications are very less. 4, Sometimes, correlation may be non- sense. Example. There may be a correlation between temperature and height of the persons, which does not means that Regression Regression means stepping back to the means value. It expresses average relationship between two or more variables. But regression analysis indicates the cause and effect relationship. The variable constituting cause is taken as independent variable and the variable constituting the effect is taken as dependent variables. Regression analysis measures linear and non-linear relationships of the variables. Hence practical applications are more. There is no non-sense regression. real relationship.Regression Analysis 5. Correlation measures the direction and degree of linear relationship. It is symmetric i.e. rxy = ryy and it is immaterial whether variables are independent or dependent. Correlation coefficient is a relative measure of the linear relationship and is independent of the units of measurement, It is pure number between —1 and +1. a PROBLEMS PROBLEM 1. Twelve pairs of observatio lines of ¥ on X and X and Y. Estimate the value of y ifx = 65. Regression analysis measures the functional relationship between the variables and it also uses to estimate the value of dependent variable for any given value of independent variable. It identifies the nature of the variable i.e. which is independent and which is dependent variable. Regression coeffi- cients are not symmetric i.e. byx # bxy The regression coefficients bxy and bxy are absolute measures. It measures change in value of one variable to change in the other variable. If the function of the regression curve is known, then the value of dependent variable can be obtain for a given value of independent variable. This value is in the units of measurement of the variable. ns are given below. Obtain the regression x | 56] 42 [72 7 36] 63] 47 | 55) 49] 98] 2] 62] 60 y | 147 | 125 | 165 | 118 | 149 | 128 | 150 | 145 | 115 | 132 | 152 | 160 SOLUTION x L vi xe ve zi 56 147 3136 | 21609 | 8232 42 125 1764 15625, 5250 72 165 5184 27225 11880 36 | 118 1296 13924 4248 63 149 3969 22201 9387 47 128 2209 16384 6016 55 150 3025, 22500 8250 49 145 2401 21025 7105 38 115 1444 13225, 4370 42 132 1764 17424 5544 62 152 3844 23104 9424 60 160 3600 | 22500 | 9000 D »=1676 | ¥ x?= 33636 | J y2=236746| J xy, = 88706 rca rca a Sicnwen a5]a Regre sat 7a hh ' n ae i y=,D } ly a | Cov (K,¥) = 5D aii“ FI 4 88706 _ (53 3) (199.67) = 153.07 2 Correlation coefficient Cov ¥) __ 153.07 ___ 9) | ne oxoy .8 x 14, | Regression line of Y on X | conor ee yeas) 24 x 14.87 10.8 (x - 51.83) y = 1.3123x + 70.65 Regression line of X on Y y — 189.67 = (0.9531) eOK eS x =o x” 10. =~ 51.88 = (0.9531) 175g - 139.67) x = 0.6922 y~ 44.85 Estimation of y If x = 65, by using regression line of Y on X J = 1.8123 x 65 + 70.65 = 155,95 PROBLEM 2. The sales and profits of a com ‘ pany are give a in regressi lines. Estimate profit if the sales of the product ic Re 72 lakhs. an eee Sales in lakhs Rs. | 75 70 Profit in lakhs Rs, 59 65Regression Analysis SOLUTION Let Xbe the sales of a product Let Y be profit of the product Let U =X~69, V=Y~60_ x vu 75 59 70: | 66 55 45 65 52 60 60 69 62 80 70 65 55 59 45 61 49 eI ou col yi a» [00 ; oy = nde ~ ©) = FP -8.8)? = 8.03 Cov (U, V) = > yvj-7.0 A” 60! = eo — (3.1) (-3.8) = 49.12 10 __ Cov(,¥) "xy =Tov= Oy oy 49.12 = (781) (6.08) = 98368 a =69, b=60, h=1, k=1Now ¥ =a+hu=69+ (3.1) = 65.9 jo y =b+kv =60-3.8 = 56.2 0x =h oy =7.31 Oy = k oy =8.03 Cov (X,Y) =h'k. Cov (U, V) = 49.12 Regression line of Y on X eI roy = -y =—*(w-x) y oy (8.03) ¥-56.2 = 0.8368 7.31 *~ 85.9) Y = 0.9192x— 4.3767 Regression line of X on Y 7.31 *—65.9 = (0.8368) (535) (- 56.2) * = 0.7618y + 23.0886 Ifx = 72, we have to estimate y ~. By using regression line of ¥ on X Y = 0.9192 x~ 4.3767 ¥ = 0.9192 (72) _ 4.3767 = 61.8057 *. If the sales of the product is lakhs, Rs. 72 lakhs,ee ion line of Y on X. To estimate marks in statistics (¥), it requires regression © a y-F =e 16.8 = _~- (x — 39.5) y-49.5 =0.42x i038” 3 y = 0.6533x + 23.6933 If marks in mathematics (x) is 52 then } = (0.6533) x 52 + 23.6933 57.6649 ~ 58 marks : ‘ .. Estimated marks in Statistics is 58 marks. (ii) Angle between two regression lines ee xox, | @ ator (oe Gd 1-(0.42)2 __ (10.8) (16.8) 5) =Tan' ("9.42 *(10.8)?+ (16.8 = Tan“ (0.892) @ =41.73° GEER CER ——— iecriea 1. Obtain regression line and estimate the value of X for a given Y = 70 for the following data. Kes | e206 2] ‘67, 67 | 68 69 |. 70__| aa i ae 72. 72 [ (69 | ‘Ans. y = 0.665x + 23.78, x = 0.54y + 30.74, % = 68.54. 2. Obtain correlation coefficient and regression lines to the following data. x_| 21] 25 | 26] 24] 22] 80] 17 | 24 | 28 [ 32 | 31 | 29 | 21 28 ¥ | 10 | 20 | 24 a1 | 21 | 24 | 18 | 22 | 19 | 30 | 27 | 26 | 19 18 Ans. 7 = 0.8534, y = 0.7007x + 4.4875, x = 1.0393y + 2.135 3. The given below are birth and death rates per 1000 population in different years of a country. Obtain regression lines. Estimate death rate of the country if birth rate is 10.5 and also estimate birth rate if death rate is 10.25, A Die 7 »| Birth | ] | rate (171) 16.5 aoe 15.2 143) 13.6) 12.9] 12.3) 11.7) 113] 11.3] 11.3 ial Death ae | 23 | eal 9.1 | 22 aa | 8.9/ 85| 9.7) 9.0 a7 9.1} 9.0} 9.2 Ans. y = 0.00415x + 8.9364, x = 1275y + 12.3228, y = 8.98, x = 13.63