*Continued from Lawsuit Regression Part Three*

This regression includes all the available explanatory variables, and now provides a coefficient for BIG5 that is negative, which implies that having a Big 5 auditor reduces the probability of a lawsuit by a small fraction. The primary concern is the extremely large p-value for the BIG5 coefficient, which implies that this variable is no longer relevant to the regression model. By conducting a Wald Test on this coefficient, we failed to reject the null hypothesis that the coefficient of BIG5 is zero.

Wald Test: | ||||

Equation: Untitled | ||||

Null Hypothesis: | C(2)=0 | |||

F-statistic | 4.67E-05 | Probability | 0.994548 |

**Removing the BIG5 variable from the regression, provides the following model:**

** **

Dependent Variable: LAWSUIT | ||||

Method: Least Squares | ||||

Date: 01/17/02 Time: 15:06 | ||||

Sample: 1 1988 | ||||

Included observations: 1988 | ||||

White Heteroskedasticity-Consistent Standard Errors & Covariance | ||||

Variable | Coefficient | Std. Error | t-Statistic | Prob. |

C | -0.073815 | 0.014354 | -5.142396 | 0.0000 |

LOGSALES | 0.012882 | 0.002756 | 4.673628 | 0.0000 |

CHEMDUM | 0.025029 | 0.016572 | 1.510333 | 0.1311 |

COMPDUM | 0.035772 | 0.013900 | 2.573440 | 0.0101 |

ELECDUM | 0.016708 | 0.013951 | 1.197599 | 0.2312 |

BETA | 0.061541 | 0.009101 | 6.762179 | 0.0000 |

CUMRET | -0.012170 | 0.003944 | -3.085824 | 0.0021 |

SIGMARET | 0.118635 | 0.074700 | 1.588162 | 0.1124 |

R-squared | 0.073388 | Mean dependent var | 0.055835 | |

Adjusted R-squared | 0.070112 | S.D. dependent var | 0.229661 | |

S.E. of regression | 0.221463 | Akaike info criterion | -0.173104 | |

Sum squared resid | 97.11108 | Schwarz criterion | -0.150589 | |

Log likelihood | 180.0650 | F-statistic | 22.40241 | |

Durbin-Watson stat | 0.103945 | Prob(F-statistic) | 0.000000 |

**Excluding BIG5 from the regression had no affect on the adjusted R-squared, and it did not change the probability of a lawsuit, on average, and with a small change to the 95% confidence interval of (-0.50,0.50).**

**4. Refining the Regression Model**

Using variables that are statistically significant gives the regression:

**Variable Significance:**

**C**: If you are not is any of the listed industries and have a 0 Beta, 0 sales and 0 cumulative returns you have a 0 increase in probability that you will get sued

** **

**Beta**: 1 beta higher increases probability by .06 that your company get sued

**Logsales**: 1% increase in log sales increases probability by .012

**Cumret**: Cannot have a negative probability, removal reduced adjusted R-squared. (Low occurrence of suits)

**Dummy Significance**:

**Chemdum**: If in Chemical industry increases probability of getting sued by .023

**Compdum**: If in Computer industry increases probability of getting sued by .035

**Electdum**: If in Electrical industry increases probability of getting sued by .016

**It seems that if you are in business in the computer industry this increases the probability of being sued by .035. This is significantly more than the other industries.**

**2. Probability of having a Lawsuit**

Already established is that using a Big5 firm does not increase a firms potential of being involved in litigation. The regressions presented shows that there is no statistical significance sowing that there is a relationship

Wald test?

P-value rule

Adjusted R-squared

However it would be interesting to understand the probability of getting sued by using equation # 3

Probability of Getting Sued in Computer Industry Given average Cumret, Beta and logsales

.07 probability of getting sued in the computer industry with average cumret, log sales and beta

95% confidence that probability will fall between .51 and -.037

**Probability of getting sued in the electrical industry given average variables**

.05 probability of getting sued in the Electrical industry

95% confidence between .49 and -.39

**Probability of getting sued in the Chemical industry**