Logistic Regression Help

#1
I am having difficulty interpreting the significance of this output:

Attainment and still enrolled (Dependent variable)

Time the job executed : 10/23/09 17:41
Full sample weight : WTA000
Variance estimation method : BRR
Type of analysis : LOGISTIC
Number of replicates : 200
Percentage of observed over total cases : 7.50%
Convergence criterion : 1.00E-06
Maximum of iterations : 25
Value of alpha (Confidence level %) : 0.05000 (95.00000 %)

Options : Intercept
Degrees of freedom : 200

Dependent variable :
No degree not enrolled 1
vs. Attained still enrolled 0

Negative log-likelihood(Pseudo R**2) 0.033
-2 log-likelihood
G**2 0 For intercept (-2LL 0) -221030.11
G**M 0 For model (-2LL M) 14622.357
Likelihood ratio (Cox-Snell) 0.042
Likelihood ratio (Cox-Snell) Maximum 0.731
Likelihood ratio (Estrella) 0.043
Degree of freedom for Wald 200
Number of categories 5
No problems detected

Filters
* Distance education courses in 2003-04=Yes. Yes" (Value= )"
Missing Values Excluded


ODDS RATIO RESULTS
Odds Ratio Lower 95%Upper 95% t Probability t Ln(Odds Ratio)
Intercept 0.203 0.107 0.387 -4.965 0 -1.592
Gender
"Male" 0.849 0.605 1.191 -0.971 0.333 -0.164
vs. "Female"
Studentís age as of 12/31/2003 1.046 1.02 1.071 3.67 0 0.045
Race/ethnicity
"Black or African American" 1.296 0.78 2.152 1.025 0.307 0.259
"Hispanic or Latino" 1.052 0.626 1.766 0.196 0.845 0.051
"Asian" 0.794 0.327 1.926 -0.524 0.601 -0.231
vs. "White"

HYPOTHESIS TESTING RESULTS
WaldF Num. DF Denom. DF Probability F
Overall Fit 3.852 5 196 0.002
Gender 0.944 1 200 0.333
Studentís age as of 12/31/2003 13.471 1 200 0
Race/ethnicity 0.476 3 198 0.699

What determines statistical significance of the model?
 
#2
Model Fit

I get the following for model fit. What determines the significance of the model?

HYPOTHESIS TESTING RESULTS
WaldF Num. DF Denom. DF Probability F
Overall Fit 3.852 5 196 0.002
Gender 0.944 1 200 0.333
Studentís age
as of 12/31/2003 13.471 1 200 0
Race/ethnicity 0.476 3 198 0.699

Is the Probability of F the statistic that indicates statistical significance for the independent covariates? Any HELP is greatly appreciated.:confused:
 

terzi

TS Contributor
#3
Logistic Regression model is estimated using maximum likelihood. For this kind of estimation, the Likelihood Ratio Test allows to compare different models hence it will allow to test significance in a model. That information can be found where the -2 Log-Likelihood values is shown. By the way, how did you get an F value? I didn't know that was possible :) I mean, maybe it is correct, I'm just curious. Which software are you using?

Maximum likelihood models are quiet different from traditional OLS Regression, so be careful when performing the analysis. Please feel free to ask when confused