Knowledge

Linear probability model

Source 📝

1200: 975: 818: 593:
Failed to parse (SVG (MathML can be enabled via browser plugin): Invalid response ("Math extension cannot connect to Restbase.") from server "http://localhost:6011/en.wikipedia.org/v1/":): {\displaystyle F_{\varepsilon|\mathbf x}(\varepsilon\mid \mathbf x) = \frac {\varepsilon +
986: 562: 829: 663: 1375: 316: 453:
More formally, the LPM can arise from a latent-variable formulation (usually to be found in the econometrics literature), as follows: assume the following regression model with a latent (unobservable) dependent variable:
1278: 1195:{\displaystyle =1-F_{\varepsilon |\mathbf {x} }(-b_{0}-\mathbf {x} '\mathbf {b} \mid \mathbf {x} )=1-{\frac {-b_{0}-\mathbf {x} '\mathbf {b} +a}{2a}}={\frac {b_{0}+a}{2a}}+{\frac {\mathbf {x} '\mathbf {b} }{2a}}.} 1543:
Horrace, William C., and Ronald L. Oaxaca. "Results on the Bias and Inconsistency of Ordinary Least Squares for the Linear Probability Model." Economics Letters, 2006: Vol. 90, P. 321–327
373: 157: 460: 1380:
This method is a general device to obtain a conditional probability model of a binary variable: if we assume that the distribution of the error term is Logistic, we obtain the
970:{\displaystyle ={\rm {Pr}}(\varepsilon >-b_{0}-\mathbf {x} '\mathbf {b} \mid \mathbf {x} )=1-{\rm {Pr}}(\varepsilon \leq -b_{0}-\mathbf {x} '\mathbf {b} \mid \mathbf {x} )} 589: 655: 813:{\displaystyle {\rm {Pr}}(y=1\mid \mathbf {x} )={\rm {Pr}}(y^{*}>0\mid \mathbf {x} )={\rm {Pr}}(b_{0}+\mathbf {x} '\mathbf {b} +\varepsilon >0\mid \mathbf {x} )} 400: 622: 88: 68: 435: 1289: 35:
for each observation takes values which are either 0 or 1. The probability of observing a 0 or 1 in any one case is treated as depending on one or more
168: 1534: 1211: 36: 32: 1556: 1476: 1515: 1492: 1437: 568: 39:. For the "linear probability model", this relationship is a particularly simple one, and allows the model to be fitted by 557:{\displaystyle y^{*}=b_{0}+\mathbf {x} '\mathbf {b} +\varepsilon ,\;\;\varepsilon \mid \mathbf {x} \sim U(-a,a).} 332: 1389: 325:. This method of fitting would be inefficient, and can be improved by adopting an iterative scheme based on 96: 329:, in which the model from the previous iteration is used to supply estimates of the conditional variances, 326: 1401: 574: 567:
The critical assumption here is that the error term of this regression is a symmetric around zero
376: 627: 1530: 1511: 1503: 1488: 1433: 375:, which would vary between observations. This approach can be related to fitting the model by 40: 28: 1525:
Wooldridge, Jeffrey M. (2013). "A Binary Dependent Variable: The Linear Probability Model".
1480: 385: 47: 601: 1370:{\displaystyle \beta _{0}={\frac {b_{0}+a}{2a}},\;\;\beta ={\frac {\mathbf {b} }{2a}}.} 73: 53: 408: 1550: 403: 322: 1385: 442: 571:
random variable, and hence, of mean zero. The cumulative distribution function of
1381: 438: 20: 1529:(5th international ed.). Mason, OH: South-Western. pp. 238–243. 1388:
and, if we assume that it is the logarithm of a Weibull distribution, the
311:{\displaystyle E=0\cdot \Pr(Y=0|X)+1\cdot \Pr(Y=1|X)=\Pr(Y=1|X)=x'\beta ,} 1273:{\displaystyle P(y=1\mid \mathbf {x} )=\beta _{0}+\mathbf {x} '\beta } 382:
A drawback of this model is that, unless restrictions are placed on
1453:
Amemiya, Takeshi (1981). "Qualitative Response Models: A Survey".
402:, the estimated coefficients can imply probabilities outside the 657:, and zero otherwise, and consider the conditional probability 321:
and hence the vector of parameters β can be estimated using
1384:, while if we assume that it is the Normal, we obtain the 70:, and its associated vector of explanatory variables, 1292: 1214: 989: 832: 666: 630: 604: 577: 463: 411: 388: 335: 171: 99: 76: 56: 1369: 1272: 1194: 969: 812: 649: 616: 583: 556: 429: 394: 367: 310: 151: 82: 62: 265: 236: 201: 100: 46:The model assumes that, for a binary outcome ( 1510:. Oxford: Basil Blackwell. pp. 267–359. 8: 1527:Introductory Econometrics: A Modern Approach 1485:Linear Probability, Logit, and Probit Models 592: 368:{\displaystyle \operatorname {Var} (Y|X=x)} 1340: 1339: 1205:But this is the Linear Probability Model, 515: 514: 1349: 1347: 1313: 1306: 1297: 1291: 1258: 1248: 1233: 1213: 1173: 1164: 1160: 1134: 1127: 1102: 1093: 1083: 1073: 1056: 1048: 1039: 1029: 1012: 1007: 1003: 988: 959: 951: 942: 932: 907: 906: 889: 881: 872: 862: 837: 836: 831: 802: 782: 773: 763: 747: 746: 735: 720: 704: 703: 692: 668: 667: 665: 635: 629: 603: 576: 522: 500: 491: 481: 468: 462: 410: 387: 348: 334: 280: 251: 216: 181: 170: 115: 98: 75: 55: 1428:Cox, D. R. (1970). "Simple Regression". 1413: 437:. For this reason, models such as the 152:{\displaystyle \Pr(Y=1|X=x)=x'\beta .} 1423: 1421: 1419: 1417: 7: 1432:. London: Methuen. pp. 33–42. 911: 908: 841: 838: 751: 748: 708: 705: 672: 669: 14: 1350: 1259: 1234: 1174: 1165: 1103: 1094: 1057: 1049: 1040: 1013: 960: 952: 943: 890: 882: 873: 803: 783: 774: 736: 693: 523: 501: 492: 1481:"The Linear Probability Model" 1455:Journal of Economic Literature 1238: 1218: 1061: 1019: 1008: 964: 916: 894: 846: 807: 756: 740: 713: 697: 677: 598:Define the indicator variable 548: 533: 424: 412: 362: 349: 342: 288: 281: 268: 259: 252: 239: 224: 217: 204: 189: 182: 175: 129: 116: 103: 1: 1504:"Qualitative Response Models" 1479:; Nelson, Forrest D. (1984). 27:(LPM) is a special case of a 584:{\displaystyle \varepsilon } 1390:complementary log-log model 449:Latent-variable formulation 1573: 650:{\displaystyle y^{*}>0} 1557:Generalized linear models 1502:Amemiya, Takeshi (1985). 445:are more commonly used. 25:linear probability model 1487:. Sage. pp. 9–29. 1430:Analysis of Binary Data 1371: 1274: 1196: 971: 814: 651: 618: 585: 558: 431: 396: 395:{\displaystyle \beta } 369: 327:weighted least squares 312: 153: 84: 64: 1508:Advanced Econometrics 1372: 1275: 1197: 972: 815: 652: 619: 586: 559: 432: 397: 370: 313: 154: 85: 65: 37:explanatory variables 1402:Linear approximation 1290: 1212: 987: 830: 664: 628: 602: 575: 461: 409: 386: 333: 169: 97: 74: 54: 617:{\displaystyle y=1} 1367: 1270: 1192: 967: 810: 647: 614: 595: 581: 554: 427: 392: 377:maximum likelihood 365: 308: 149: 80: 60: 33:dependent variable 1536:978-1-111-53439-4 1362: 1334: 1283:with the mapping 1187: 1155: 1122: 83:{\displaystyle X} 63:{\displaystyle Y} 41:linear regression 29:binary regression 1564: 1540: 1521: 1498: 1477:Aldrich, John H. 1463: 1462: 1450: 1444: 1443: 1425: 1376: 1374: 1373: 1368: 1363: 1361: 1353: 1348: 1335: 1333: 1325: 1318: 1317: 1307: 1302: 1301: 1279: 1277: 1276: 1271: 1266: 1262: 1253: 1252: 1237: 1201: 1199: 1198: 1193: 1188: 1186: 1178: 1177: 1172: 1168: 1161: 1156: 1154: 1146: 1139: 1138: 1128: 1123: 1121: 1113: 1106: 1101: 1097: 1088: 1087: 1074: 1060: 1052: 1047: 1043: 1034: 1033: 1018: 1017: 1016: 1011: 976: 974: 973: 968: 963: 955: 950: 946: 937: 936: 915: 914: 893: 885: 880: 876: 867: 866: 845: 844: 819: 817: 816: 811: 806: 786: 781: 777: 768: 767: 755: 754: 739: 725: 724: 712: 711: 696: 676: 675: 656: 654: 653: 648: 640: 639: 623: 621: 620: 615: 590: 588: 587: 582: 563: 561: 560: 555: 526: 504: 499: 495: 486: 485: 473: 472: 436: 434: 433: 430:{\displaystyle } 428: 401: 399: 398: 393: 374: 372: 371: 366: 352: 317: 315: 314: 309: 301: 284: 255: 220: 185: 162:For this model, 158: 156: 155: 150: 142: 119: 89: 87: 86: 81: 69: 67: 66: 61: 31:model. Here the 16:Statistics model 1572: 1571: 1567: 1566: 1565: 1563: 1562: 1561: 1547: 1546: 1537: 1524: 1518: 1501: 1495: 1475: 1472: 1470:Further reading 1467: 1466: 1461:(4): 1483–1536. 1452: 1451: 1447: 1440: 1427: 1426: 1415: 1410: 1398: 1354: 1326: 1309: 1308: 1293: 1288: 1287: 1257: 1244: 1210: 1209: 1179: 1163: 1162: 1147: 1130: 1129: 1114: 1092: 1079: 1075: 1038: 1025: 999: 985: 984: 980: 941: 928: 871: 858: 828: 827: 823: 772: 759: 716: 662: 661: 631: 626: 625: 600: 599: 573: 572: 490: 477: 464: 459: 458: 451: 407: 406: 384: 383: 331: 330: 294: 167: 166: 135: 95: 94: 72: 71: 52: 51: 48:Bernoulli trial 17: 12: 11: 5: 1570: 1568: 1560: 1559: 1549: 1548: 1545: 1544: 1541: 1535: 1522: 1516: 1499: 1493: 1471: 1468: 1465: 1464: 1445: 1438: 1412: 1411: 1409: 1406: 1405: 1404: 1397: 1394: 1378: 1377: 1366: 1360: 1357: 1352: 1346: 1343: 1338: 1332: 1329: 1324: 1321: 1316: 1312: 1305: 1300: 1296: 1281: 1280: 1269: 1265: 1261: 1256: 1251: 1247: 1243: 1240: 1236: 1232: 1229: 1226: 1223: 1220: 1217: 1203: 1202: 1191: 1185: 1182: 1176: 1171: 1167: 1159: 1153: 1150: 1145: 1142: 1137: 1133: 1126: 1120: 1117: 1112: 1109: 1105: 1100: 1096: 1091: 1086: 1082: 1078: 1072: 1069: 1066: 1063: 1059: 1055: 1051: 1046: 1042: 1037: 1032: 1028: 1024: 1021: 1015: 1010: 1006: 1002: 998: 995: 992: 978: 977: 966: 962: 958: 954: 949: 945: 940: 935: 931: 927: 924: 921: 918: 913: 910: 905: 902: 899: 896: 892: 888: 884: 879: 875: 870: 865: 861: 857: 854: 851: 848: 843: 840: 835: 821: 820: 809: 805: 801: 798: 795: 792: 789: 785: 780: 776: 771: 766: 762: 758: 753: 750: 745: 742: 738: 734: 731: 728: 723: 719: 715: 710: 707: 702: 699: 695: 691: 688: 685: 682: 679: 674: 671: 646: 643: 638: 634: 613: 610: 607: 580: 565: 564: 553: 550: 547: 544: 541: 538: 535: 532: 529: 525: 521: 518: 513: 510: 507: 503: 498: 494: 489: 484: 480: 476: 471: 467: 450: 447: 426: 423: 420: 417: 414: 391: 364: 361: 358: 355: 351: 347: 344: 341: 338: 319: 318: 307: 304: 300: 297: 293: 290: 287: 283: 279: 276: 273: 270: 267: 264: 261: 258: 254: 250: 247: 244: 241: 238: 235: 232: 229: 226: 223: 219: 215: 212: 209: 206: 203: 200: 197: 194: 191: 188: 184: 180: 177: 174: 160: 159: 148: 145: 141: 138: 134: 131: 128: 125: 122: 118: 114: 111: 108: 105: 102: 79: 59: 15: 13: 10: 9: 6: 4: 3: 2: 1569: 1558: 1555: 1554: 1552: 1542: 1538: 1532: 1528: 1523: 1519: 1517:0-631-13345-3 1513: 1509: 1505: 1500: 1496: 1494:0-8039-2133-0 1490: 1486: 1482: 1478: 1474: 1473: 1469: 1460: 1456: 1449: 1446: 1441: 1439:0-416-10400-2 1435: 1431: 1424: 1422: 1420: 1418: 1414: 1407: 1403: 1400: 1399: 1395: 1393: 1391: 1387: 1383: 1364: 1358: 1355: 1344: 1341: 1336: 1330: 1327: 1322: 1319: 1314: 1310: 1303: 1298: 1294: 1286: 1285: 1284: 1267: 1263: 1254: 1249: 1245: 1241: 1230: 1227: 1224: 1221: 1215: 1208: 1207: 1206: 1189: 1183: 1180: 1169: 1157: 1151: 1148: 1143: 1140: 1135: 1131: 1124: 1118: 1115: 1110: 1107: 1098: 1089: 1084: 1080: 1076: 1070: 1067: 1064: 1053: 1044: 1035: 1030: 1026: 1022: 1004: 1000: 996: 993: 990: 983: 982: 981: 956: 947: 938: 933: 929: 925: 922: 919: 903: 900: 897: 886: 877: 868: 863: 859: 855: 852: 849: 833: 826: 825: 824: 799: 796: 793: 790: 787: 778: 769: 764: 760: 743: 732: 729: 726: 721: 717: 700: 689: 686: 683: 680: 660: 659: 658: 644: 641: 636: 632: 611: 608: 605: 596: 578: 570: 551: 545: 542: 539: 536: 530: 527: 519: 516: 511: 508: 505: 496: 487: 482: 478: 474: 469: 465: 457: 456: 455: 448: 446: 444: 440: 421: 418: 415: 405: 404:unit interval 389: 380: 378: 359: 356: 353: 345: 339: 336: 328: 324: 323:least squares 305: 302: 298: 295: 291: 285: 277: 274: 271: 262: 256: 248: 245: 242: 233: 230: 227: 221: 213: 210: 207: 198: 195: 192: 186: 178: 172: 165: 164: 163: 146: 143: 139: 136: 132: 126: 123: 120: 112: 109: 106: 93: 92: 91: 77: 57: 49: 44: 42: 38: 34: 30: 26: 22: 1526: 1507: 1484: 1458: 1454: 1448: 1429: 1386:probit model 1379: 1282: 1204: 979: 822: 597: 566: 452: 443:probit model 381: 320: 161: 45: 24: 18: 1382:logit model 439:logit model 1408:References 21:statistics 1342:β 1295:β 1268:β 1246:β 1231:∣ 1090:− 1077:− 1071:− 1054:∣ 1036:− 1023:− 1005:ε 997:− 957:∣ 939:− 926:− 923:≤ 920:ε 904:− 887:∣ 869:− 856:− 850:ε 800:∣ 791:ε 733:∣ 722:∗ 690:∣ 637:∗ 579:ε 537:− 528:∼ 520:∣ 517:ε 509:ε 470:∗ 390:β 340:⁡ 303:β 234:⋅ 199:⋅ 144:β 1551:Category 1396:See also 1264:′ 1170:′ 1099:′ 1045:′ 948:′ 878:′ 779:′ 594:a}{2a}.} 591:here is 497:′ 299:′ 140:′ 569:uniform 441:or the 1533:  1514:  1491:  1436:  1531:ISBN 1512:ISBN 1489:ISBN 1434:ISBN 853:> 794:> 727:> 642:> 23:, a 624:if 337:Var 50:), 19:In 1553:: 1506:. 1483:. 1459:19 1457:. 1416:^ 1392:. 379:. 266:Pr 237:Pr 202:Pr 101:Pr 90:, 43:. 1539:. 1520:. 1497:. 1442:. 1365:. 1359:a 1356:2 1351:b 1345:= 1337:, 1331:a 1328:2 1323:a 1320:+ 1315:0 1311:b 1304:= 1299:0 1260:x 1255:+ 1250:0 1242:= 1239:) 1235:x 1228:1 1225:= 1222:y 1219:( 1216:P 1190:. 1184:a 1181:2 1175:b 1166:x 1158:+ 1152:a 1149:2 1144:a 1141:+ 1136:0 1132:b 1125:= 1119:a 1116:2 1111:a 1108:+ 1104:b 1095:x 1085:0 1081:b 1068:1 1065:= 1062:) 1058:x 1050:b 1041:x 1031:0 1027:b 1020:( 1014:x 1009:| 1001:F 994:1 991:= 965:) 961:x 953:b 944:x 934:0 930:b 917:( 912:r 909:P 901:1 898:= 895:) 891:x 883:b 874:x 864:0 860:b 847:( 842:r 839:P 834:= 808:) 804:x 797:0 788:+ 784:b 775:x 770:+ 765:0 761:b 757:( 752:r 749:P 744:= 741:) 737:x 730:0 718:y 714:( 709:r 706:P 701:= 698:) 694:x 687:1 684:= 681:y 678:( 673:r 670:P 645:0 633:y 612:1 609:= 606:y 552:. 549:) 546:a 543:, 540:a 534:( 531:U 524:x 512:, 506:+ 502:b 493:x 488:+ 483:0 479:b 475:= 466:y 425:] 422:1 419:, 416:0 413:[ 363:) 360:x 357:= 354:X 350:| 346:Y 343:( 306:, 296:x 292:= 289:) 286:X 282:| 278:1 275:= 272:Y 269:( 263:= 260:) 257:X 253:| 249:1 246:= 243:Y 240:( 231:1 228:+ 225:) 222:X 218:| 214:0 211:= 208:Y 205:( 196:0 193:= 190:] 187:X 183:| 179:Y 176:[ 173:E 147:. 137:x 133:= 130:) 127:x 124:= 121:X 117:| 113:1 110:= 107:Y 104:( 78:X 58:Y

Index

statistics
binary regression
dependent variable
explanatory variables
linear regression
Bernoulli trial
least squares
weighted least squares
maximum likelihood
unit interval
logit model
probit model
uniform
logit model
probit model
complementary log-log model
Linear approximation




ISBN
0-416-10400-2
Aldrich, John H.
"The Linear Probability Model"
ISBN
0-8039-2133-0
"Qualitative Response Models"
ISBN
0-631-13345-3

Text is available under the Creative Commons Attribution-ShareAlike License. Additional terms may apply.