Binary Regression
Six Sigma – iSixSigma › Forums › Old Forums › General › Binary Regression
 This topic has 5 replies, 3 voices, and was last updated 16 years, 10 months ago by Deep.

AuthorPosts

October 3, 2005 at 12:38 pm #40900
I was hoping to get some assistance with the interpretation of binary logistic regression.
I have two factors, one continuous and the other is discrete. Both are significant in terms of the response. I understand the interpretation of the odds ratio for the continuous factor (for each additional unit it increases the percent of my response by 9%). I’m not quite sure of the interpretation of the odds ratio for the discrete factor which is .27.
Is it also possible to build an equation for both factors for predicting the response? This would be synonomous with writing the equation for a continuous response?
Thanks for any advice.0October 3, 2005 at 12:46 pm #127786
Ken FeldmanParticipant@Darth Include @Darth in your post and this person will
be notified via email.Are you properly using the tool? From what I understand, Binary logistic requires a discrete Y and a continuous X. Didn’t see mention of the use of a discrete X.
0October 3, 2005 at 12:52 pm #127788My response is discrete. Minitab will allow you to include both continuous and discrete factors.
0October 3, 2005 at 3:36 pm #127798AB:
You said>>continuous factor (for each additional unit it increases the percent of my response by 9%).
I dont know how did you come up with this statement? Did you make this statement because you got an odd ratio of .09 for your continuous variable? If that is the case then your statement (interpretation) is not correct. For interpreting this correctly we need the reference level of your response.
Other information needed to answer this question, what is your Link function (You can see this at the top the sessions window). What are the reference levels, of both Y and Xs.
It would be good if you can copy and paste the sessions window here.
Thanks
Deep0October 5, 2005 at 2:23 pm #127888Here is my session window. Both factors are discrete with two categories.
Link Function: Logit
Response Information
Variable Value Count
ams client yes 37 (Event)
no 1745
Total 1782
Logistic Regression Table
Odds 95% CI
Predictor Coef SE Coef Z P Ratio Lower Upper
Constant 3.82748 0.201771 18.97 0.000
pfp sent
Y 1.65808 0.729752 2.27 0.023 0.19 0.05 0.80
Funded Cat
>15 0.784532 0.366535 2.14 0.032 2.19 1.07 4.49
LogLikelihood = 173.818
Test that all slopes are zero: G = 12.309, DF = 2, PValue = 0.002
GoodnessofFit Tests
Method ChiSquare DF P
Pearson 0.268379 1 0.604
Deviance 0.253976 1 0.614
HosmerLemeshow 0.113079 1 0.737
Table of Observed and Expected Frequencies:
(See HosmerLemeshow Test for the Pearson ChiSquare Statistic)
Value yes no
Group Observed Expected Observed Expected Total
1 1 1.3 322 321.7 323
2 26 25.3 1206 1206.7 1232
3 10 10.3 217 216.7 227
Measures of Association:
(Between the Response Variable and Predicted Probabilities)
Pairs Number Percent Summary Measures
Concordant 25477 39.5 Somers’ D 0.27
Discordant 8198 12.7 GoodmanKruskal Gamma 0.51
Ties 30890 47.8 Kendall’s Taua 0.01
Total 64565 100.0
0October 5, 2005 at 3:16 pm #127892AB:
Thanks for pasting the sessions window.
You might read /interpret the Odds ratio like this
An AMS client who has Funded cat>15 has an Odds of 2.19 times larger of saying Yes than who has <15 Funded cat (On the assumption that the reference level for funded cat is 15.
More AMS client who have pfp sent Y has said No. In other words, the AMS client who have pfp Y said Yes is only 19% when compared to AMS client said Yes having pfp N (On the assumption that the reference level for pfp is N)
Practically if you need AMS client Yes you might go with pfp N
Hope this helps. Please let me know; also please consider English is not my first language. Besides I dont know what these factors mean in practical :0.
Deep0 
AuthorPosts
The forum ‘General’ is closed to new topics and replies.