**************************************************************** * * Dichotomous dependent variable example * * Bert Kritzer * March 22, 2004 * **************************************************************** drop _all use "D:\COURSES\PS551\anes2000\KRITZE~1.DTA" * prepare data drop if v000361==. gen FT_bush=v000361 drop if v000523>6 gen PID=v000523 gen gender=(v001029==2) gen vote=(v001249==3) if v001249<4 * linear probability model regress vote FT_bush PID gender * discriminate function discrim vote FT_bush PID gender,predict * logistic regression logistic vote FT_bush PID gender logit lstat * probit probit vote FT_bush PID gender dprobit vote FT_bush PID gender probit . **************************************************************** . * . * Dichotomous dependent variable example . * . * Bert Kritzer . * March 22, 2004 . * . **************************************************************** . . . drop _all . use "D:\COURSES\PS551\anes2000\KRITZE~1.DTA" . . * prepare data . drop if v000361==. (36 observations deleted) . gen FT_bush=v000361 . drop if v000523>6 (23 observations deleted) . gen PID=v000523 . gen gender=(v001029==2) . gen vote=(v001249==3) if v001249<4 (393 missing values generated) . . * linear probability model . regress vote FT_bush PID gender Source | SS df MS Number of obs = 1103 -------------+------------------------------ F( 3, 1099) = 673.58 Model | 178.22933 3 59.4097767 Prob > F = 0.0000 Residual | 96.9311413 1099 .088199401 R-squared = 0.6477 -------------+------------------------------ Adj R-squared = 0.6468 Total | 275.160471 1102 .249691898 Root MSE = .29698 ------------------------------------------------------------------------------ vote | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- FT_bush | .0056604 .0004498 12.59 0.000 .0047779 .0065429 PID | .1321134 .0052579 25.13 0.000 .1217968 .14243 gender | -.0334144 .018079 -1.85 0.065 -.0688877 .0020589 _cons | -.202717 .0243498 -8.33 0.000 -.2504944 -.1549396 ------------------------------------------------------------------------------ . . * discriminate function . discrim vote FT_bush PID gender,predict Dichotomous Discriminant Analysis Observations = 1103 Obs Group 0 = 577 Indep variables = 3 Obs Group 1 = 526 Centroid 0 = -1.2935 R-square = 0.6477 Centroid 1 = 1.4189 Mahalanobis = 7.3573 Grand Cntd = 0.1254 Eigenvalue = 1.8387 Wilk's Lambda = 0.3523 Canon. Corr. = 0.8048 Chi-square = 1.1e+03 Eta Squared = 0.6477 Sign Chi2 = 0.0000 Discrim Function Unstandardized Variable Coefficients Coefficients ------------------------------------------------- FT_bush -0.0643 0.0237 PID -1.5006 0.5532 gender 0.3795 -0.1399 constant 7.8894 -2.8459 ----- Predicted ----- Actual | Group 0 Group 1 | Total ---------+--------------------------+-------- Group 0 | 521 56 | 577 Group 1 | 56 470 | 526 ---------+--------------------------+-------- Total | 577 526 | 1103 ---------+--------------------------+-------- Correctly predicted = 89.85 % Model sensitivity = 90.29 % Model specificity = 89.35 % False positive = 9.71 % False negative = 10.65 % . . * logistic regression . logistic vote FT_bush PID gender Logistic regression Number of obs = 1103 LR chi2(3) = 953.07 Prob > chi2 = 0.0000 Log likelihood = -286.82608 Pseudo R2 = 0.6243 ------------------------------------------------------------------------------ vote | Odds Ratio Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------- FT_bush | 1.06158 .0061795 10.27 0.000 1.049537 1.073761 PID | 2.619961 .1752123 14.40 0.000 2.298105 2.986893 gender | .763331 .1667252 -1.24 0.216 .4975024 1.171199 ------------------------------------------------------------------------------ . logit Logit estimates Number of obs = 1103 LR chi2(3) = 953.07 Prob > chi2 = 0.0000 Log likelihood = -286.82608 Pseudo R2 = 0.6243 ------------------------------------------------------------------------------ vote | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------- FT_bush | .059758 .005821 10.27 0.000 .048349 .071167 PID | .9631594 .0668759 14.40 0.000 .832085 1.094234 gender | -.2700635 .218418 -1.24 0.216 -.698155 .1580279 _cons | -6.206523 .4265949 -14.55 0.000 -7.042633 -5.370412 ------------------------------------------------------------------------------ . lstat Logistic model for vote -------- True -------- Classified | D ~D | Total -----------+--------------------------+----------- + | 463 52 | 515 - | 63 525 | 588 -----------+--------------------------+----------- Total | 526 577 | 1103 Classified + if predicted Pr(D) >= .5 True D defined as vote != 0 -------------------------------------------------- Sensitivity Pr( +| D) 88.02% Specificity Pr( -|~D) 90.99% Positive predictive value Pr( D| +) 89.90% Negative predictive value Pr(~D| -) 89.29% -------------------------------------------------- False + rate for true ~D Pr( +|~D) 9.01% False - rate for true D Pr( -| D) 11.98% False + rate for classified + Pr(~D| +) 10.10% False - rate for classified - Pr( D| -) 10.71% -------------------------------------------------- Correctly classified 89.57% -------------------------------------------------- . . * probit . probit vote FT_bush PID gender Iteration 0: log likelihood = -763.36186 Iteration 1: log likelihood = -349.31102 Iteration 2: log likelihood = -294.2439 Iteration 3: log likelihood = -287.84513 Iteration 4: log likelihood = -287.68735 Iteration 5: log likelihood = -287.68723 Probit estimates Number of obs = 1103 LR chi2(3) = 951.35 Prob > chi2 = 0.0000 Log likelihood = -287.68723 Pseudo R2 = 0.6231 ------------------------------------------------------------------------------ vote | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------- FT_bush | .0325355 .003059 10.64 0.000 .0265401 .038531 PID | .5362596 .0340104 15.77 0.000 .4696005 .6029188 gender | -.1452686 .1176437 -1.23 0.217 -.3758459 .0853088 _cons | -3.42494 .2128225 -16.09 0.000 -3.842064 -3.007816 ------------------------------------------------------------------------------ . dprobit vote FT_bush PID gender Iteration 0: log likelihood = -763.36186 Iteration 1: log likelihood = -349.31102 Iteration 2: log likelihood = -294.2439 Iteration 3: log likelihood = -287.84513 Iteration 4: log likelihood = -287.68735 Iteration 5: log likelihood = -287.68723 Probit estimates Number of obs = 1103 LR chi2(3) = 951.35 Prob > chi2 = 0.0000 Log likelihood = -287.68723 Pseudo R2 = 0.6231 ------------------------------------------------------------------------------ vote | dF/dx Std. Err. z P>|z| x-bar [ 95% C.I. ] ---------+-------------------------------------------------------------------- FT_bush | .0128816 .0012049 10.64 0.000 57.3799 .01052 .015243 PID | .2123185 .0135625 15.77 0.000 2.82593 .185736 .238901 gender*| -.0575202 .046549 -1.23 0.217 .55485 -.148755 .033714 ---------+-------------------------------------------------------------------- obs. P | .4768812 pred. P | .4509643 (at x-bar) ------------------------------------------------------------------------------ (*) dF/dx is for discrete change of dummy variable from 0 to 1 z and P>|z| are the test of the underlying coefficient being 0 . probit Probit estimates Number of obs = 1103 LR chi2(3) = 951.35 Prob > chi2 = 0.0000 Log likelihood = -287.68723 Pseudo R2 = 0.6231 ------------------------------------------------------------------------------ vote | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------- FT_bush | .0325355 .003059 10.64 0.000 .0265401 .038531 PID | .5362596 .0340104 15.77 0.000 .4696005 .6029188 gender | -.1452686 .1176437 -1.23 0.217 -.3758459 .0853088 _cons | -3.42494 .2128225 -16.09 0.000 -3.842064 -3.007816 ------------------------------------------------------------------------------ . . end of do-file . log close log: D:\COURSES\PS552\Stata Do Files\LogisticRegression.log log type: text closed on: 24 Mar 2004, 16:01:03 -------------------------------------------------------------------------------