Probit and Logit Models with Normal Errors

​
dataset index
1
method
linear
probit
logit
confidence level
0.95
ANOVA table
best fit
parameter table
scale for confidence ellipsoid
small
medium
large
data and fit
confidence ellipsoid
ANOVATable
DF
SS
MS
Model
2
27.7741
13.8871
Error
48
1.79011
0.037294
Uncorrected Total
50
29.5642
​
Corrected Total
49
15.7879
​
BestFit
0.680535-0.23218x
ParameterTable
Estimate
Standard Error
t-Statistic
P-Value
a
-0.23218
0.0119844
-19.3736
2.47429×
-24
10
b
0.680535
0.0284677
23.9055
2.61262×
-28
10
On many occasions, it is appropriate to use a simple linear model to regress data. On other occasions, however, such as when the dependent variable is a probability, transformed linear combinations of the independent variables so that their values are contained within the interval[0,1]. This Demonstration takes 10 sample datasets and compares a simple linear regression to two frequently used alternatives: the probit model and the logit model. The fitting used assumes normally distributed residuals. You select the dataset, the regression model you wish to examine, and the set of regression report items you wish to see.

Details

In some instances, the confidence ellipsoid will be too small to see.

External Links

Fitting Generalized Linear Models (Wolfram Library Archive)
Logit Transformation (Wolfram MathWorld)
Regression (Wolfram MathWorld)

Permanent Citation

Seth J. Chandler
​
​"Probit and Logit Models with Normal Errors"​
​http://demonstrations.wolfram.com/ProbitAndLogitModelsWithNormalErrors/​
​Wolfram Demonstrations Project​
​Published: October 29, 2008