logistic regression - james · pdf filegoals of this section 2/ 8 speci c goals: learn how to...
TRANSCRIPT
![Page 1: Logistic Regression - James · PDF fileGoals of this section 2/ 8 Speci c goals: Learn how to conduct regression analysis with a dummy ... Logistic Regression: method for estimating](https://reader035.vdocuments.us/reader035/viewer/2022062600/5a78b4c27f8b9a21538c3140/html5/thumbnails/1.jpg)
Logistic Regression
BUS 735: Business Decision Making and Research
BUS 735: Business Decision Making and Research Logistic Regression
![Page 2: Logistic Regression - James · PDF fileGoals of this section 2/ 8 Speci c goals: Learn how to conduct regression analysis with a dummy ... Logistic Regression: method for estimating](https://reader035.vdocuments.us/reader035/viewer/2022062600/5a78b4c27f8b9a21538c3140/html5/thumbnails/2.jpg)
Goals of this section 2/ 8
Specific goals:
Learn how to conduct regression analysis with a dummyindependent variable.
Learning objectives:
LO2: Be able to construct and use multiple regression models(including some limited dependent variable models) toconstruct and test hypotheses considering complexrelationships among multiple variables.LO6: Be able to use standard computer packages such asSPSS and Excel to conduct the quantitative analyses describedin the learning objectives above.LO7: Have a sound familiarity of various statistical andquantitative methods in order to be able to approach abusiness decision problem and be able to select appropriatemethods to answer the question.
BUS 735: Business Decision Making and Research Logistic Regression
![Page 3: Logistic Regression - James · PDF fileGoals of this section 2/ 8 Speci c goals: Learn how to conduct regression analysis with a dummy ... Logistic Regression: method for estimating](https://reader035.vdocuments.us/reader035/viewer/2022062600/5a78b4c27f8b9a21538c3140/html5/thumbnails/3.jpg)
Goals of this section 2/ 8
Specific goals:
Learn how to conduct regression analysis with a dummyindependent variable.
Learning objectives:
LO2: Be able to construct and use multiple regression models(including some limited dependent variable models) toconstruct and test hypotheses considering complexrelationships among multiple variables.LO6: Be able to use standard computer packages such asSPSS and Excel to conduct the quantitative analyses describedin the learning objectives above.LO7: Have a sound familiarity of various statistical andquantitative methods in order to be able to approach abusiness decision problem and be able to select appropriatemethods to answer the question.
BUS 735: Business Decision Making and Research Logistic Regression
![Page 4: Logistic Regression - James · PDF fileGoals of this section 2/ 8 Speci c goals: Learn how to conduct regression analysis with a dummy ... Logistic Regression: method for estimating](https://reader035.vdocuments.us/reader035/viewer/2022062600/5a78b4c27f8b9a21538c3140/html5/thumbnails/4.jpg)
Logistic Regression 3/ 8
Logistic Regression: method for estimating a regression witha dummy dependent variable.
Will a potential customer purchase a product?(YES=1, NO=0).
Might use explanatory variables: age, gender, income, etc.
Will a potential employee be retained after one year?(YES=1, NO=0).
Might use explanatory variables: age, gender, years experience,past income, education dummy (4-year = 1, otherwise = 0).
Why not run a regression? Which assumption is violated?
Dependent variable is not interval. As a result, residual willnot be normally distributed.
BUS 735: Business Decision Making and Research Logistic Regression
![Page 5: Logistic Regression - James · PDF fileGoals of this section 2/ 8 Speci c goals: Learn how to conduct regression analysis with a dummy ... Logistic Regression: method for estimating](https://reader035.vdocuments.us/reader035/viewer/2022062600/5a78b4c27f8b9a21538c3140/html5/thumbnails/5.jpg)
Logistic Regression 3/ 8
Logistic Regression: method for estimating a regression witha dummy dependent variable.
Will a potential customer purchase a product?(YES=1, NO=0).
Might use explanatory variables: age, gender, income, etc.
Will a potential employee be retained after one year?(YES=1, NO=0).
Might use explanatory variables: age, gender, years experience,past income, education dummy (4-year = 1, otherwise = 0).
Why not run a regression? Which assumption is violated?
Dependent variable is not interval. As a result, residual willnot be normally distributed.
BUS 735: Business Decision Making and Research Logistic Regression
![Page 6: Logistic Regression - James · PDF fileGoals of this section 2/ 8 Speci c goals: Learn how to conduct regression analysis with a dummy ... Logistic Regression: method for estimating](https://reader035.vdocuments.us/reader035/viewer/2022062600/5a78b4c27f8b9a21538c3140/html5/thumbnails/6.jpg)
Logistic Regression 3/ 8
Logistic Regression: method for estimating a regression witha dummy dependent variable.
Will a potential customer purchase a product?(YES=1, NO=0).
Might use explanatory variables: age, gender, income, etc.
Will a potential employee be retained after one year?(YES=1, NO=0).
Might use explanatory variables: age, gender, years experience,past income, education dummy (4-year = 1, otherwise = 0).
Why not run a regression? Which assumption is violated?
Dependent variable is not interval. As a result, residual willnot be normally distributed.
BUS 735: Business Decision Making and Research Logistic Regression
![Page 7: Logistic Regression - James · PDF fileGoals of this section 2/ 8 Speci c goals: Learn how to conduct regression analysis with a dummy ... Logistic Regression: method for estimating](https://reader035.vdocuments.us/reader035/viewer/2022062600/5a78b4c27f8b9a21538c3140/html5/thumbnails/7.jpg)
Logistic Regression 3/ 8
Logistic Regression: method for estimating a regression witha dummy dependent variable.
Will a potential customer purchase a product?(YES=1, NO=0).
Might use explanatory variables: age, gender, income, etc.
Will a potential employee be retained after one year?(YES=1, NO=0).
Might use explanatory variables: age, gender, years experience,past income, education dummy (4-year = 1, otherwise = 0).
Why not run a regression? Which assumption is violated?
Dependent variable is not interval. As a result, residual willnot be normally distributed.
BUS 735: Business Decision Making and Research Logistic Regression
![Page 8: Logistic Regression - James · PDF fileGoals of this section 2/ 8 Speci c goals: Learn how to conduct regression analysis with a dummy ... Logistic Regression: method for estimating](https://reader035.vdocuments.us/reader035/viewer/2022062600/5a78b4c27f8b9a21538c3140/html5/thumbnails/8.jpg)
Logistic Regression 3/ 8
Logistic Regression: method for estimating a regression witha dummy dependent variable.
Will a potential customer purchase a product?(YES=1, NO=0).
Might use explanatory variables: age, gender, income, etc.
Will a potential employee be retained after one year?(YES=1, NO=0).
Might use explanatory variables: age, gender, years experience,past income, education dummy (4-year = 1, otherwise = 0).
Why not run a regression? Which assumption is violated?
Dependent variable is not interval. As a result, residual willnot be normally distributed.
BUS 735: Business Decision Making and Research Logistic Regression
![Page 9: Logistic Regression - James · PDF fileGoals of this section 2/ 8 Speci c goals: Learn how to conduct regression analysis with a dummy ... Logistic Regression: method for estimating](https://reader035.vdocuments.us/reader035/viewer/2022062600/5a78b4c27f8b9a21538c3140/html5/thumbnails/9.jpg)
Logistic Regression 3/ 8
Logistic Regression: method for estimating a regression witha dummy dependent variable.
Will a potential customer purchase a product?(YES=1, NO=0).
Might use explanatory variables: age, gender, income, etc.
Will a potential employee be retained after one year?(YES=1, NO=0).
Might use explanatory variables: age, gender, years experience,past income, education dummy (4-year = 1, otherwise = 0).
Why not run a regression? Which assumption is violated?Dependent variable is not interval. As a result, residual willnot be normally distributed.
BUS 735: Business Decision Making and Research Logistic Regression
![Page 10: Logistic Regression - James · PDF fileGoals of this section 2/ 8 Speci c goals: Learn how to conduct regression analysis with a dummy ... Logistic Regression: method for estimating](https://reader035.vdocuments.us/reader035/viewer/2022062600/5a78b4c27f8b9a21538c3140/html5/thumbnails/10.jpg)
Structural Form 4/ 8
Normal regression:
yi = b0 + b1X1,i + b2X2,i + ... + bk−1Xk−1,i + ei
Logistic regression:
log (Odds) = b0 + b1X1,i + b2X2,i + ... + bk−1Xk−1,i + ei
Odds =P(yi = 1)
1− P(yi = 1)=
P(yi = 1)
P(yi = 0)
Remember, yi ∈ 0, 1 indicates YES=1 event did occur, orNO=0 event did not occur.
BUS 735: Business Decision Making and Research Logistic Regression
![Page 11: Logistic Regression - James · PDF fileGoals of this section 2/ 8 Speci c goals: Learn how to conduct regression analysis with a dummy ... Logistic Regression: method for estimating](https://reader035.vdocuments.us/reader035/viewer/2022062600/5a78b4c27f8b9a21538c3140/html5/thumbnails/11.jpg)
Structural Form 4/ 8
Normal regression:
yi = b0 + b1X1,i + b2X2,i + ... + bk−1Xk−1,i + ei
Logistic regression:
log (Odds) = b0 + b1X1,i + b2X2,i + ... + bk−1Xk−1,i + ei
Odds =P(yi = 1)
1− P(yi = 1)=
P(yi = 1)
P(yi = 0)
Remember, yi ∈ 0, 1 indicates YES=1 event did occur, orNO=0 event did not occur.
BUS 735: Business Decision Making and Research Logistic Regression
![Page 12: Logistic Regression - James · PDF fileGoals of this section 2/ 8 Speci c goals: Learn how to conduct regression analysis with a dummy ... Logistic Regression: method for estimating](https://reader035.vdocuments.us/reader035/viewer/2022062600/5a78b4c27f8b9a21538c3140/html5/thumbnails/12.jpg)
Structural Form 4/ 8
Normal regression:
yi = b0 + b1X1,i + b2X2,i + ... + bk−1Xk−1,i + ei
Logistic regression:
log (Odds) = b0 + b1X1,i + b2X2,i + ... + bk−1Xk−1,i + ei
Odds =P(yi = 1)
1− P(yi = 1)=
P(yi = 1)
P(yi = 0)
Remember, yi ∈ 0, 1 indicates YES=1 event did occur, orNO=0 event did not occur.
BUS 735: Business Decision Making and Research Logistic Regression
![Page 13: Logistic Regression - James · PDF fileGoals of this section 2/ 8 Speci c goals: Learn how to conduct regression analysis with a dummy ... Logistic Regression: method for estimating](https://reader035.vdocuments.us/reader035/viewer/2022062600/5a78b4c27f8b9a21538c3140/html5/thumbnails/13.jpg)
Structural Form 4/ 8
Normal regression:
yi = b0 + b1X1,i + b2X2,i + ... + bk−1Xk−1,i + ei
Logistic regression:
log (Odds) = b0 + b1X1,i + b2X2,i + ... + bk−1Xk−1,i + ei
Odds =P(yi = 1)
1− P(yi = 1)=
P(yi = 1)
P(yi = 0)
Remember, yi ∈ 0, 1 indicates YES=1 event did occur, orNO=0 event did not occur.
BUS 735: Business Decision Making and Research Logistic Regression
![Page 14: Logistic Regression - James · PDF fileGoals of this section 2/ 8 Speci c goals: Learn how to conduct regression analysis with a dummy ... Logistic Regression: method for estimating](https://reader035.vdocuments.us/reader035/viewer/2022062600/5a78b4c27f8b9a21538c3140/html5/thumbnails/14.jpg)
Structural Form 4/ 8
Normal regression:
yi = b0 + b1X1,i + b2X2,i + ... + bk−1Xk−1,i + ei
Logistic regression:
log (Odds) = b0 + b1X1,i + b2X2,i + ... + bk−1Xk−1,i + ei
Odds =P(yi = 1)
1− P(yi = 1)=
P(yi = 1)
P(yi = 0)
Remember, yi ∈ 0, 1 indicates YES=1 event did occur, orNO=0 event did not occur.
BUS 735: Business Decision Making and Research Logistic Regression
![Page 15: Logistic Regression - James · PDF fileGoals of this section 2/ 8 Speci c goals: Learn how to conduct regression analysis with a dummy ... Logistic Regression: method for estimating](https://reader035.vdocuments.us/reader035/viewer/2022062600/5a78b4c27f8b9a21538c3140/html5/thumbnails/15.jpg)
Structural Form 4/ 8
Normal regression:
yi = b0 + b1X1,i + b2X2,i + ... + bk−1Xk−1,i + ei
Logistic regression:
log (Odds) = b0 + b1X1,i + b2X2,i + ... + bk−1Xk−1,i + ei
Odds =P(yi = 1)
1− P(yi = 1)=
P(yi = 1)
P(yi = 0)
Remember, yi ∈ 0, 1 indicates YES=1 event did occur, orNO=0 event did not occur.
BUS 735: Business Decision Making and Research Logistic Regression
![Page 16: Logistic Regression - James · PDF fileGoals of this section 2/ 8 Speci c goals: Learn how to conduct regression analysis with a dummy ... Logistic Regression: method for estimating](https://reader035.vdocuments.us/reader035/viewer/2022062600/5a78b4c27f8b9a21538c3140/html5/thumbnails/16.jpg)
Predicted Value 5/ 8
Predicted value from a regular regression:
yi = b0 + b1X1,i + b2X2,i + ... + bk−1Xk−1,i
For a logistic regression, you can get predicted logist (not toointeresting yet):
li = ln(Odds) = b0 + b1X1,i + b2X2,i + ... + bk−1Xk−1,i
To uncover the predicted probability of the eventoccuring:
P(yi = 1) =1
1 + e(−li )
BUS 735: Business Decision Making and Research Logistic Regression
![Page 17: Logistic Regression - James · PDF fileGoals of this section 2/ 8 Speci c goals: Learn how to conduct regression analysis with a dummy ... Logistic Regression: method for estimating](https://reader035.vdocuments.us/reader035/viewer/2022062600/5a78b4c27f8b9a21538c3140/html5/thumbnails/17.jpg)
Predicted Value 5/ 8
Predicted value from a regular regression:
yi = b0 + b1X1,i + b2X2,i + ... + bk−1Xk−1,i
For a logistic regression, you can get predicted logist (not toointeresting yet):
li = ln(Odds) = b0 + b1X1,i + b2X2,i + ... + bk−1Xk−1,i
To uncover the predicted probability of the eventoccuring:
P(yi = 1) =1
1 + e(−li )
BUS 735: Business Decision Making and Research Logistic Regression
![Page 18: Logistic Regression - James · PDF fileGoals of this section 2/ 8 Speci c goals: Learn how to conduct regression analysis with a dummy ... Logistic Regression: method for estimating](https://reader035.vdocuments.us/reader035/viewer/2022062600/5a78b4c27f8b9a21538c3140/html5/thumbnails/18.jpg)
Predicted Value 5/ 8
Predicted value from a regular regression:
yi = b0 + b1X1,i + b2X2,i + ... + bk−1Xk−1,i
For a logistic regression, you can get predicted logist (not toointeresting yet):
li = ln(Odds) = b0 + b1X1,i + b2X2,i + ... + bk−1Xk−1,i
To uncover the predicted probability of the eventoccuring:
P(yi = 1) =1
1 + e(−li )
BUS 735: Business Decision Making and Research Logistic Regression
![Page 19: Logistic Regression - James · PDF fileGoals of this section 2/ 8 Speci c goals: Learn how to conduct regression analysis with a dummy ... Logistic Regression: method for estimating](https://reader035.vdocuments.us/reader035/viewer/2022062600/5a78b4c27f8b9a21538c3140/html5/thumbnails/19.jpg)
Predicted Value 5/ 8
Predicted value from a regular regression:
yi = b0 + b1X1,i + b2X2,i + ... + bk−1Xk−1,i
For a logistic regression, you can get predicted logist (not toointeresting yet):
li = ln(Odds) = b0 + b1X1,i + b2X2,i + ... + bk−1Xk−1,i
To uncover the predicted probability of the eventoccuring:
P(yi = 1) =1
1 + e(−li )
BUS 735: Business Decision Making and Research Logistic Regression
![Page 20: Logistic Regression - James · PDF fileGoals of this section 2/ 8 Speci c goals: Learn how to conduct regression analysis with a dummy ... Logistic Regression: method for estimating](https://reader035.vdocuments.us/reader035/viewer/2022062600/5a78b4c27f8b9a21538c3140/html5/thumbnails/20.jpg)
Predicted Value 5/ 8
Predicted value from a regular regression:
yi = b0 + b1X1,i + b2X2,i + ... + bk−1Xk−1,i
For a logistic regression, you can get predicted logist (not toointeresting yet):
li = ln(Odds) = b0 + b1X1,i + b2X2,i + ... + bk−1Xk−1,i
To uncover the predicted probability of the eventoccuring:
P(yi = 1) =1
1 + e(−li )
BUS 735: Business Decision Making and Research Logistic Regression
![Page 21: Logistic Regression - James · PDF fileGoals of this section 2/ 8 Speci c goals: Learn how to conduct regression analysis with a dummy ... Logistic Regression: method for estimating](https://reader035.vdocuments.us/reader035/viewer/2022062600/5a78b4c27f8b9a21538c3140/html5/thumbnails/21.jpg)
Predicted Value 5/ 8
Predicted value from a regular regression:
yi = b0 + b1X1,i + b2X2,i + ... + bk−1Xk−1,i
For a logistic regression, you can get predicted logist (not toointeresting yet):
li = ln(Odds) = b0 + b1X1,i + b2X2,i + ... + bk−1Xk−1,i
To uncover the predicted probability of the eventoccuring:
P(yi = 1) =1
1 + e(−li )
BUS 735: Business Decision Making and Research Logistic Regression
![Page 22: Logistic Regression - James · PDF fileGoals of this section 2/ 8 Speci c goals: Learn how to conduct regression analysis with a dummy ... Logistic Regression: method for estimating](https://reader035.vdocuments.us/reader035/viewer/2022062600/5a78b4c27f8b9a21538c3140/html5/thumbnails/22.jpg)
Marginal Effects 6/ 8
Marginal effect for regression: measure of how much ychanges when x increases by 1.
Example: How much does public expenditure per capitaincrease (or decrease) when economic ability increases by oneunit?
Marginal effect for logit: measure of how much P(yi = 1)changes when x increases by 1.
How much more (or less) likely will an interview candicate beworking here after one year if she/he has a four year collegeeducation?
BUS 735: Business Decision Making and Research Logistic Regression
![Page 23: Logistic Regression - James · PDF fileGoals of this section 2/ 8 Speci c goals: Learn how to conduct regression analysis with a dummy ... Logistic Regression: method for estimating](https://reader035.vdocuments.us/reader035/viewer/2022062600/5a78b4c27f8b9a21538c3140/html5/thumbnails/23.jpg)
Marginal Effects 6/ 8
Marginal effect for regression: measure of how much ychanges when x increases by 1.
Example: How much does public expenditure per capitaincrease (or decrease) when economic ability increases by oneunit?
Marginal effect for logit: measure of how much P(yi = 1)changes when x increases by 1.
How much more (or less) likely will an interview candicate beworking here after one year if she/he has a four year collegeeducation?
BUS 735: Business Decision Making and Research Logistic Regression
![Page 24: Logistic Regression - James · PDF fileGoals of this section 2/ 8 Speci c goals: Learn how to conduct regression analysis with a dummy ... Logistic Regression: method for estimating](https://reader035.vdocuments.us/reader035/viewer/2022062600/5a78b4c27f8b9a21538c3140/html5/thumbnails/24.jpg)
Marginal Effects 6/ 8
Marginal effect for regression: measure of how much ychanges when x increases by 1.
Example: How much does public expenditure per capitaincrease (or decrease) when economic ability increases by oneunit?
Marginal effect for logit: measure of how much P(yi = 1)changes when x increases by 1.
How much more (or less) likely will an interview candicate beworking here after one year if she/he has a four year collegeeducation?
BUS 735: Business Decision Making and Research Logistic Regression
![Page 25: Logistic Regression - James · PDF fileGoals of this section 2/ 8 Speci c goals: Learn how to conduct regression analysis with a dummy ... Logistic Regression: method for estimating](https://reader035.vdocuments.us/reader035/viewer/2022062600/5a78b4c27f8b9a21538c3140/html5/thumbnails/25.jpg)
Marginal Effects 6/ 8
Marginal effect for regression: measure of how much ychanges when x increases by 1.
Example: How much does public expenditure per capitaincrease (or decrease) when economic ability increases by oneunit?
Marginal effect for logit: measure of how much P(yi = 1)changes when x increases by 1.
How much more (or less) likely will an interview candicate beworking here after one year if she/he has a four year collegeeducation?
BUS 735: Business Decision Making and Research Logistic Regression
![Page 26: Logistic Regression - James · PDF fileGoals of this section 2/ 8 Speci c goals: Learn how to conduct regression analysis with a dummy ... Logistic Regression: method for estimating](https://reader035.vdocuments.us/reader035/viewer/2022062600/5a78b4c27f8b9a21538c3140/html5/thumbnails/26.jpg)
Coefficients versus Marginal Effects 7/ 8
For a regular regression, for coefficient b2:
The sign (positive/negative) indicates whether x2 causes y toincrease or decrease.The magnitude tells how much y increases when increasing x2by 1.Coefficient = Marginal Effect.
For a logistic regression, the coefficient b2:
The sign (positive/negative) indicates whether x2 causes y toincrease or decrease.The magnitude is pretty meaningless.Need to do more to figure out marginal effect.
Because SPSS is stupid, it cannot figure out marginal effects.
For a specific individual (specific values for X’s), computepredicted probabilities by hand for:
The individual’s actual set of values for X’s.Same set of X’s, except increase one of them by 1.Take the difference = Marginal effect.
BUS 735: Business Decision Making and Research Logistic Regression
![Page 27: Logistic Regression - James · PDF fileGoals of this section 2/ 8 Speci c goals: Learn how to conduct regression analysis with a dummy ... Logistic Regression: method for estimating](https://reader035.vdocuments.us/reader035/viewer/2022062600/5a78b4c27f8b9a21538c3140/html5/thumbnails/27.jpg)
Coefficients versus Marginal Effects 7/ 8
For a regular regression, for coefficient b2:
The sign (positive/negative) indicates whether x2 causes y toincrease or decrease.The magnitude tells how much y increases when increasing x2by 1.Coefficient = Marginal Effect.
For a logistic regression, the coefficient b2:
The sign (positive/negative) indicates whether x2 causes y toincrease or decrease.The magnitude is pretty meaningless.Need to do more to figure out marginal effect.
Because SPSS is stupid, it cannot figure out marginal effects.
For a specific individual (specific values for X’s), computepredicted probabilities by hand for:
The individual’s actual set of values for X’s.Same set of X’s, except increase one of them by 1.Take the difference = Marginal effect.
BUS 735: Business Decision Making and Research Logistic Regression
![Page 28: Logistic Regression - James · PDF fileGoals of this section 2/ 8 Speci c goals: Learn how to conduct regression analysis with a dummy ... Logistic Regression: method for estimating](https://reader035.vdocuments.us/reader035/viewer/2022062600/5a78b4c27f8b9a21538c3140/html5/thumbnails/28.jpg)
Coefficients versus Marginal Effects 7/ 8
For a regular regression, for coefficient b2:
The sign (positive/negative) indicates whether x2 causes y toincrease or decrease.The magnitude tells how much y increases when increasing x2by 1.Coefficient = Marginal Effect.
For a logistic regression, the coefficient b2:
The sign (positive/negative) indicates whether x2 causes y toincrease or decrease.The magnitude is pretty meaningless.Need to do more to figure out marginal effect.
Because SPSS is stupid, it cannot figure out marginal effects.
For a specific individual (specific values for X’s), computepredicted probabilities by hand for:
The individual’s actual set of values for X’s.Same set of X’s, except increase one of them by 1.Take the difference = Marginal effect.
BUS 735: Business Decision Making and Research Logistic Regression
![Page 29: Logistic Regression - James · PDF fileGoals of this section 2/ 8 Speci c goals: Learn how to conduct regression analysis with a dummy ... Logistic Regression: method for estimating](https://reader035.vdocuments.us/reader035/viewer/2022062600/5a78b4c27f8b9a21538c3140/html5/thumbnails/29.jpg)
Coefficients versus Marginal Effects 7/ 8
For a regular regression, for coefficient b2:
The sign (positive/negative) indicates whether x2 causes y toincrease or decrease.The magnitude tells how much y increases when increasing x2by 1.Coefficient = Marginal Effect.
For a logistic regression, the coefficient b2:
The sign (positive/negative) indicates whether x2 causes y toincrease or decrease.The magnitude is pretty meaningless.Need to do more to figure out marginal effect.
Because SPSS is stupid, it cannot figure out marginal effects.
For a specific individual (specific values for X’s), computepredicted probabilities by hand for:
The individual’s actual set of values for X’s.Same set of X’s, except increase one of them by 1.Take the difference = Marginal effect.
BUS 735: Business Decision Making and Research Logistic Regression
![Page 30: Logistic Regression - James · PDF fileGoals of this section 2/ 8 Speci c goals: Learn how to conduct regression analysis with a dummy ... Logistic Regression: method for estimating](https://reader035.vdocuments.us/reader035/viewer/2022062600/5a78b4c27f8b9a21538c3140/html5/thumbnails/30.jpg)
Coefficients versus Marginal Effects 7/ 8
For a regular regression, for coefficient b2:
The sign (positive/negative) indicates whether x2 causes y toincrease or decrease.The magnitude tells how much y increases when increasing x2by 1.Coefficient = Marginal Effect.
For a logistic regression, the coefficient b2:
The sign (positive/negative) indicates whether x2 causes y toincrease or decrease.The magnitude is pretty meaningless.Need to do more to figure out marginal effect.
Because SPSS is stupid, it cannot figure out marginal effects.
For a specific individual (specific values for X’s), computepredicted probabilities by hand for:
The individual’s actual set of values for X’s.Same set of X’s, except increase one of them by 1.Take the difference = Marginal effect.
BUS 735: Business Decision Making and Research Logistic Regression
![Page 31: Logistic Regression - James · PDF fileGoals of this section 2/ 8 Speci c goals: Learn how to conduct regression analysis with a dummy ... Logistic Regression: method for estimating](https://reader035.vdocuments.us/reader035/viewer/2022062600/5a78b4c27f8b9a21538c3140/html5/thumbnails/31.jpg)
Coefficients versus Marginal Effects 7/ 8
For a regular regression, for coefficient b2:
The sign (positive/negative) indicates whether x2 causes y toincrease or decrease.The magnitude tells how much y increases when increasing x2by 1.Coefficient = Marginal Effect.
For a logistic regression, the coefficient b2:
The sign (positive/negative) indicates whether x2 causes y toincrease or decrease.The magnitude is pretty meaningless.Need to do more to figure out marginal effect.
Because SPSS is stupid, it cannot figure out marginal effects.
For a specific individual (specific values for X’s), computepredicted probabilities by hand for:
The individual’s actual set of values for X’s.Same set of X’s, except increase one of them by 1.Take the difference = Marginal effect.
BUS 735: Business Decision Making and Research Logistic Regression
![Page 32: Logistic Regression - James · PDF fileGoals of this section 2/ 8 Speci c goals: Learn how to conduct regression analysis with a dummy ... Logistic Regression: method for estimating](https://reader035.vdocuments.us/reader035/viewer/2022062600/5a78b4c27f8b9a21538c3140/html5/thumbnails/32.jpg)
Coefficients versus Marginal Effects 7/ 8
For a regular regression, for coefficient b2:
The sign (positive/negative) indicates whether x2 causes y toincrease or decrease.The magnitude tells how much y increases when increasing x2by 1.Coefficient = Marginal Effect.
For a logistic regression, the coefficient b2:
The sign (positive/negative) indicates whether x2 causes y toincrease or decrease.The magnitude is pretty meaningless.Need to do more to figure out marginal effect.
Because SPSS is stupid, it cannot figure out marginal effects.
For a specific individual (specific values for X’s), computepredicted probabilities by hand for:
The individual’s actual set of values for X’s.Same set of X’s, except increase one of them by 1.Take the difference = Marginal effect.
BUS 735: Business Decision Making and Research Logistic Regression
![Page 33: Logistic Regression - James · PDF fileGoals of this section 2/ 8 Speci c goals: Learn how to conduct regression analysis with a dummy ... Logistic Regression: method for estimating](https://reader035.vdocuments.us/reader035/viewer/2022062600/5a78b4c27f8b9a21538c3140/html5/thumbnails/33.jpg)
Coefficients versus Marginal Effects 7/ 8
For a regular regression, for coefficient b2:
The sign (positive/negative) indicates whether x2 causes y toincrease or decrease.The magnitude tells how much y increases when increasing x2by 1.Coefficient = Marginal Effect.
For a logistic regression, the coefficient b2:
The sign (positive/negative) indicates whether x2 causes y toincrease or decrease.The magnitude is pretty meaningless.Need to do more to figure out marginal effect.
Because SPSS is stupid, it cannot figure out marginal effects.
For a specific individual (specific values for X’s), computepredicted probabilities by hand for:
The individual’s actual set of values for X’s.Same set of X’s, except increase one of them by 1.Take the difference = Marginal effect.
BUS 735: Business Decision Making and Research Logistic Regression
![Page 34: Logistic Regression - James · PDF fileGoals of this section 2/ 8 Speci c goals: Learn how to conduct regression analysis with a dummy ... Logistic Regression: method for estimating](https://reader035.vdocuments.us/reader035/viewer/2022062600/5a78b4c27f8b9a21538c3140/html5/thumbnails/34.jpg)
Coefficients versus Marginal Effects 7/ 8
For a regular regression, for coefficient b2:
The sign (positive/negative) indicates whether x2 causes y toincrease or decrease.The magnitude tells how much y increases when increasing x2by 1.Coefficient = Marginal Effect.
For a logistic regression, the coefficient b2:
The sign (positive/negative) indicates whether x2 causes y toincrease or decrease.The magnitude is pretty meaningless.Need to do more to figure out marginal effect.
Because SPSS is stupid, it cannot figure out marginal effects.
For a specific individual (specific values for X’s), computepredicted probabilities by hand for:
The individual’s actual set of values for X’s.Same set of X’s, except increase one of them by 1.Take the difference = Marginal effect.
BUS 735: Business Decision Making and Research Logistic Regression
![Page 35: Logistic Regression - James · PDF fileGoals of this section 2/ 8 Speci c goals: Learn how to conduct regression analysis with a dummy ... Logistic Regression: method for estimating](https://reader035.vdocuments.us/reader035/viewer/2022062600/5a78b4c27f8b9a21538c3140/html5/thumbnails/35.jpg)
Coefficients versus Marginal Effects 7/ 8
For a regular regression, for coefficient b2:
The sign (positive/negative) indicates whether x2 causes y toincrease or decrease.The magnitude tells how much y increases when increasing x2by 1.Coefficient = Marginal Effect.
For a logistic regression, the coefficient b2:
The sign (positive/negative) indicates whether x2 causes y toincrease or decrease.The magnitude is pretty meaningless.Need to do more to figure out marginal effect.
Because SPSS is stupid, it cannot figure out marginal effects.
For a specific individual (specific values for X’s), computepredicted probabilities by hand for:
The individual’s actual set of values for X’s.Same set of X’s, except increase one of them by 1.Take the difference = Marginal effect.
BUS 735: Business Decision Making and Research Logistic Regression
![Page 36: Logistic Regression - James · PDF fileGoals of this section 2/ 8 Speci c goals: Learn how to conduct regression analysis with a dummy ... Logistic Regression: method for estimating](https://reader035.vdocuments.us/reader035/viewer/2022062600/5a78b4c27f8b9a21538c3140/html5/thumbnails/36.jpg)
Coefficients versus Marginal Effects 7/ 8
For a regular regression, for coefficient b2:
The sign (positive/negative) indicates whether x2 causes y toincrease or decrease.The magnitude tells how much y increases when increasing x2by 1.Coefficient = Marginal Effect.
For a logistic regression, the coefficient b2:
The sign (positive/negative) indicates whether x2 causes y toincrease or decrease.The magnitude is pretty meaningless.Need to do more to figure out marginal effect.
Because SPSS is stupid, it cannot figure out marginal effects.
For a specific individual (specific values for X’s), computepredicted probabilities by hand for:
The individual’s actual set of values for X’s.Same set of X’s, except increase one of them by 1.Take the difference = Marginal effect.
BUS 735: Business Decision Making and Research Logistic Regression
![Page 37: Logistic Regression - James · PDF fileGoals of this section 2/ 8 Speci c goals: Learn how to conduct regression analysis with a dummy ... Logistic Regression: method for estimating](https://reader035.vdocuments.us/reader035/viewer/2022062600/5a78b4c27f8b9a21538c3140/html5/thumbnails/37.jpg)
Coefficients versus Marginal Effects 7/ 8
For a regular regression, for coefficient b2:
The sign (positive/negative) indicates whether x2 causes y toincrease or decrease.The magnitude tells how much y increases when increasing x2by 1.Coefficient = Marginal Effect.
For a logistic regression, the coefficient b2:
The sign (positive/negative) indicates whether x2 causes y toincrease or decrease.The magnitude is pretty meaningless.Need to do more to figure out marginal effect.
Because SPSS is stupid, it cannot figure out marginal effects.
For a specific individual (specific values for X’s), computepredicted probabilities by hand for:
The individual’s actual set of values for X’s.Same set of X’s, except increase one of them by 1.Take the difference = Marginal effect.
BUS 735: Business Decision Making and Research Logistic Regression
![Page 38: Logistic Regression - James · PDF fileGoals of this section 2/ 8 Speci c goals: Learn how to conduct regression analysis with a dummy ... Logistic Regression: method for estimating](https://reader035.vdocuments.us/reader035/viewer/2022062600/5a78b4c27f8b9a21538c3140/html5/thumbnails/38.jpg)
Coefficients versus Marginal Effects 7/ 8
For a regular regression, for coefficient b2:
The sign (positive/negative) indicates whether x2 causes y toincrease or decrease.The magnitude tells how much y increases when increasing x2by 1.Coefficient = Marginal Effect.
For a logistic regression, the coefficient b2:
The sign (positive/negative) indicates whether x2 causes y toincrease or decrease.The magnitude is pretty meaningless.Need to do more to figure out marginal effect.
Because SPSS is stupid, it cannot figure out marginal effects.
For a specific individual (specific values for X’s), computepredicted probabilities by hand for:
The individual’s actual set of values for X’s.Same set of X’s, except increase one of them by 1.Take the difference = Marginal effect.
BUS 735: Business Decision Making and Research Logistic Regression
![Page 39: Logistic Regression - James · PDF fileGoals of this section 2/ 8 Speci c goals: Learn how to conduct regression analysis with a dummy ... Logistic Regression: method for estimating](https://reader035.vdocuments.us/reader035/viewer/2022062600/5a78b4c27f8b9a21538c3140/html5/thumbnails/39.jpg)
Example 8/ 8
Well Being Dataset (Described on pages 356-357 of SAB).
182 college students in Washington, D.C. area.
Variables of interest:
Weight (4 categories: UNDERWEIGHT=1, NORMAL=2,OVERWEIGHT=3, OBESE=4)Gender (2 categories: MALE=0, FEMALE=1)AgeRace (AFRICAN AMERICAN=1, WHITE=2, OTHER=3)Marital Status (SINGLE=1, MARRIED=2, OTHER=3)Financial Status (OVEREXTENDED=1, MAKING ENDSMEET=2, COMFORTABLE=3)Physical Health (5 point Likert scale)Depression Scale (0-60, values > 15 indicate depression)
Find preliminary evidence whether each variable is related toweight.
Logistic Regression Analysis.
BUS 735: Business Decision Making and Research Logistic Regression