Logistic Regression Chansun Hong & Colleen Orihill COM 531 Cleveland State University 4/7/2009 1
2
Fictional Scenario: CSU catering services ambitiously plans to expand its business in downtown Cleveland and needs to hire additional student employees. 274 student applications have been submitted for the limited number of catering positions. The CSU Hiring Manager is selecting students who best exude a playful warmth and generosity while connecting with the customers. He wants students who view him or herself as entertaining. With so many students applying for the limited number of catering positions, the CSU Hiring Manager needs a concrete method to identify what demographic, personality, or communication traits will predict the newly hired students to be consistently entertaining. He decided to ask the CSU School of Communication for assistance in identifying appropriate characteristics that would best predict the probability of a student entertainer. The Communication graduate students working on this research project decided to use the multivariate statistic Logistic Regression to predict the optimal student employees. Variables from the HUMGROUP 2001 Measures were utilized for the calculations. Research Question: Which demographic, personality and communication style characteristics are most likely to predict the probability of an entertaining person? Dependent Variable – Binary : B12. I have considered being an entertainer (RECODED so that true = 1; false = 0) Independent Variables (w/ measurement scale): BLOCK 1 E1. Are you male or female? (RECODED for femaleness; newvar = female as 1; male as 0 ) E2. Age (years) BLOCK 2 C8. Do you prefer action to planning for action? (Likert Scale: 1‐5) C12. Do you have frequent ups and downs in mood, either with or without apparent cause? (Likert Scale: 1‐5) C13. Would you be very unhappy if you were prevented from making social contacts? (Likert Scale: 1‐5) BLOCK 3 D6. To be friendly, I habitually acknowledge verbally other’s contributions D16. Under pressure I come across as a relaxed speaker. D19. I always find it very easy to communicate on a one‐to‐one basis with strangers. D38. I am always an extremely friendly communicator. 3
Recoding E1 Gender using the IF COMMAND FILE NEW SYNTAX (SPSS Syntax Editor appears) In the SPSS Syntax Editor, TYPE your recode using the IF Command. In this case the recode command is: IF (E1=1) newvar=0. IF (E1=2) newvar=1. NOTE: All SPSS Syntax Editor commands must end with a period [.]. The independent variable GENDER was RECODED for femaleness; newvar = female as 1; male as 0 ) RUN ALL The recode is complete and the new variable is created when: • “SPSS Processor is ready” appears on the bottom of the SPSS Syntax Editor box. • The new variable appears at the bottom of the variable list. MINIMIZE the SPSS Syntax Editor. 4
Recoding DV: 5
Logistic Regression – Enter DV and BLOCKS of IV variables ANALYZE REGRESSION BINARY LOGISTIC (Logistic Regression box appears) Enter DV: On the variable list, highlight your DV. (In this case the DV is B12.) Click the arrow to bring it into the box labeled DEPENDENT. (Notice the blued phrase “Block 1 of 1” within the Logistic Regression box, just below the DEPENDENT variable) Enter IVs – BLOCK 1: On the variable list, highlight your first IV for BLOCK 1. (In this case newvar, gender). Click the arrow to bring the highlighted IV into the box labeled COVARIATES. On the variable list, highlight your next IV for BLOCK 1. (In this case E2 Age). Click the arrow to bring the highlighted IV into the box labeled COVARIATES. METHOD: Enter NEXT BLOCK 1 is complete (and ready to create another BLOCK) when the blued phrase “Block 2 of 2” appears within the Logistic Regression box. Enter IV – BLOCK 2 & 3: The same method as Enter IVs‐BLOCK 1. 6
Logistic Regression – Select OPTIONS At the bottom of the Logistic Regression box, CLICK the button titled OPTIONS. (Logistic Regression: Options box appears) STATISTIC and PLOTS: CLASSIFICATION PLOTS HOSMER‐LEMESHOW GOODNESS OF FIT CORRELATIONS OF ESTIMATES CL FOR EXP(B) = 95% CONTINUE OPTIONAL NOTE: STATISTICS AND PLOTS: CASEWISE LISTING OF RESIDUALS ALL CASES 7
Logistic Regression – SAVE & RUN At the bottom of the Logistic Regression box, CLICK the button titled SAVE. (Logistic Regression: Save box appears) PREDICTED VALUES: PROBABILITIES RESIDUALS: UNSTANDARDIZED STANDARDIZED INFLUENCE: (do not check anything; not utilizing) CONTINUE (Logistic Regression box) OK 8
GET
FILE='C:\Documents and Settings\HyeonJoo Cheon\My Documents\COM531\ppt\dataset.sav'.
DATASET NAME DataSet1 WINDOW=FRONT.
LOGISTIC REGRESSION VARIABLES NewDV
/METHOD = ENTER newvar e2 /METHOD = ENTER c8 c12 c13 /METHOD = ENTER d6
d16 d19 d38
/SAVE = PRED RESID ZRESID
/CLASSPLOT
/PRINT = GOODFIT CORR CI(95)
/CRITERIA = PIN(.05) POUT(.10) ITERATE(20) CUT(.5) .
Logistic Regression
Notes
Output Created
01-Apr-2009 09:58:07
Comments
Input
Missing Value Handling
Data
C:\Documents and
Settings\HyeonJoo Cheon\My
Documents\COM531\ppt\dataset.
sav
Active Dataset
DataSet1
Filter
<none>
Weight
<none>
Split File
<none>
N of Rows in Working
Data File
Definition of Missing
Syntax
Resources
314
User-defined missing values are
treated as missing
LOGISTIC REGRESSION
VARIABLES NewDV
/METHOD = ENTER newvar e2
/METHOD = ENTER c8 c12 c13
/METHOD = ENTER d6
d16 d19 d38
/SAVE = PRED RESID ZRESID
/CLASSPLOT
/PRINT = GOODFIT CORR CI(95)
/CRITERIA = PIN(.05) POUT(.10)
ITERATE(20) CUT(.5) .
Processor Time
00:00:00.125
Elapsed Time
Variables Created or …
PRE_3
00:00:00.139
Predicted probability
Notes
Variables Created or
Modified
RES_3
Difference between observed and
predicted probabilities
ZRE_3
Normalized residual
Page 1
9
[DataSet1] C:\Documents and Settings\HyeonJoo Cheon\My Documents\COM531\ppt\dataset.sav
Case Processing Summary
Unweighted Cases a
Selected Cases
Included in Analysis
N
Missing Cases
Total
Unselected Cases
Total
Percent
274
87.3
40
12.7
314
100.0
0
.0
314
100.0
a. If weight is in effect, see classification table for the total number of cases.
Dependent Variable Encoding
Ori…
0
Internal Value
0
1
1
Block 0: Beginning Block
Classification Tablea,b
Predicted
being an entertainer
Step 0
Observed
being an entertainer
0
Percentage
Correct
1
0
146
0
1
128
0
Overall Percentage
100.0
.0
53.3
a. Constant is included in the model.
b. The cut value is .500
Variables in the Equation
B
Step 0
Constant
-.132
S.E.
.121
Wald
1.181
df
Sig.
1
.277
Exp(B)
.877
Page 2
10
Variables not in the Equation
Score
Step 0
Variables
newvar
Sig.
1
.010
.433
1
.511
7.479
2
.024
e2
Overall Statistics
df
6.591
Block 1: Method = Enter
Omnibus Tests of Model Coefficients
Chi-square
Step 1
df
Sig.
Step
7.528
2
.023
Block
7.528
2
.023
Model
7.528
2
.023
Model Summary
Step
1
-2 Log
likelihood
Cox & Snell R
Square
Nagelkerke R
Square
.027
.036
371.133 a
a. Estimation terminated at iteration number 3 because parameter estimates
changed by less than .001.
Hosmer and Lemeshow Test
Step
1
Chi-square
df
3.120
Sig.
8
.927
Contingency Table for Hosmer and Lemeshow Test
being an entertainer = .00
Observed
Step 1
Expected
being an entertainer = 1.00
Observed
Expected
Total
1
25
22.204
10
12.796
35
2
15
17.572
13
10.428
28
3
19
19.184
12
11.816
31
4
16
15.639
10
10.361
26
5
6
7.026
7
5.974
13
6
18
18.461
21
20.539
39
7
11
12.137
15
13.863
26
8
10
8.746
9
10.254
19
9
11
11.345
14
13.655
25
10
15
13.687
17
18.313
32
Page 3
11
Classification Tablea
Predicted
being an entertainer
Step 1
Observed
being an entertainer
0
Percentage
Correct
1
0
81
65
55.5
1
50
78
60.9
Overall Percentage
58.0
a. The cut value is .500
Variables in the Equation
95.0% C.I.for EXP(B)
B
Step
1a
newvar
e2
Constant
S.E.
Wald
df
Sig.
Exp(B)
Lower
Upper
-.655
.248
6.989
1
.008
.520
.320
.844
.026
.028
.902
1
.342
1.027
.972
1.084
-.366
.590
.384
1
.536
.694
a. Variable(s) entered on step 1: newvar, e2.
Correlation Matrix
Constant
Step 1
newvar
e2
Constant
1.000
-.080
newvar
-.080
1.000
-.123
e2
-.958
-.123
1.000
-.958
Page 4
12
Step number: 1
Observed Groups and Predicted Probabilities
80 ?
?
?
?
?
?
F
?
?
R
60 ?
?
E
?
?
Q
?
?
U
?
1
1
?
E
40 ?
1
11
?
N
?
11
11
?
C
?
11
11
?
Y
?
00
111
?
20 ?
00 1
001
?
?
0011
0011
?
?
0010
0000
?
?
0000111
0000 0
?
Predicted ????????????????????????????????????????????????????????????????????????????????????????????????????
Prob:
0
.1
.2
.3
.4
.5
.6
.7
.8
.9
1
Group: 0000000000000000000000000000000000000000000000000011111111111111111111111111111111111111111111111111
Predicted Probability is of Membership for 1.00
The Cut Value is .50
Symbols: 0 - .00
Prob:
Group:
0
.1
.2
.3
.4
.5
.6
.7
.8
.9
1
0000000000000000000000000000000000000000000000000011111111111111111111111111111111111111111111111111
Predicted Probability is of Membership for 1.00
The Cut Value is .50
Symbols: 0 - .00
1 - 1.00
Each Symbol Represents 5 Cases.
Block 2: Method = Enter
Omnibus Tests of Model Coefficients
Chi-square
Step 1
df
Sig.
Step
5.746
3
.125
Block
5.746
3
.125
Model
13.274
5
.021
Page 5
13
Model Summary
Step
1
-2 Log
likelihood
Cox & Snell R
Square
Nagelkerke R
Square
.047
.063
365.387 a
a. Estimation terminated at iteration number 3 because parameter estimates
changed by less than .001.
Hosmer and Lemeshow Test
Step
1
Chi-square
df
6.892
Sig.
8
.548
Contingency Table for Hosmer and Lemeshow Test
being an entertainer = .00
Observed
Step 1
being an entertainer = 1.00
Expected
Observed
Expected
Total
1
17
19.471
10
7.529
27
2
23
17.721
4
9.279
27
3
16
16.668
11
10.332
27
4
16
15.798
11
11.202
27
5
14
14.966
13
12.034
27
6
13
13.970
14
13.030
27
7
11
12.811
16
14.189
27
8
12
11.985
15
15.015
27
9
11
11.182
16
15.818
27
10
13
11.426
18
19.574
31
Classification Tablea
Predicted
being an entertainer
Step 1
Observed
being an entertainer
0
Percentage
Correct
1
0
98
48
67.1
1
59
69
53.9
Overall Percentage
60.9
a. The cut value is .500
Variables in the Equation
95.0% C.I.for EXP(B)
B
Step
1a
newvar
S.E.
Wald
df
Sig.
Exp(B)
Lower
Upper
-.702
.252
7.788
1
.005
.496
.303
.811
e2
.020
.028
.505
1
.477
1.020
.966
1.078
c8
.035
.052
.442
1
.506
1.035
.934
1.147
c12
.032
.043
.552
1
.458
1.032
.949
1.122
a. Variable(s) entered on step 1: c8, c12, c13.
Page 6
14
Variables in the Equation
95.0% C.I.for EXP(B)
B
Step
1a
c13
Constant
S.E.
Wald
df
Sig.
Exp(B)
.101
.049
4.238
1
.040
1.106
-1.381
.753
3.366
1
.067
.251
Lower
1.005
Upper
1.218
a. Variable(s) entered on step 1: c8, c12, c13.
Correlation Matrix
Constant
Step 1
newvar
e2
c8
-.687
-.299
1.000
-.111
-.111
1.000
-.299
-.016
-.091
-.284
.005
-.040
-.415
-.102
-.047
Constant
1.000
-.017
newvar
-.017
e2
-.687
c8
c12
c13
c12
c13
-.284
-.415
-.016
.005
-.102
-.091
-.040
-.047
1.000
-.085
-.120
-.085
1.000
.080
-.120
.080
1.000
Page 7
15
Step number: 1
Observed Groups and Predicted Probabilities
16 ?
?
?
?
?
?
F
?
1
?
R
12 ?
1
1
?
E
?
1 1
1 1
?
Q
?
1 1 1 11
111
?
U
?
1 1 1 111
1
1111
?
E
8 ?
1
1 0 1 011 1 11 11111 1
?
N
?
1 1 1 011 011 1 11 11111111
?
C
?
0 00 0 011 01111 11 1101111111 1
?
Y
?
0 0010 000100110 111 1100011111 1
?
4 ?
1
10000001000100110 101 11000011011 1
?
?
1 0110000000010001000001100 00000011010 1
?
?
10011000000000000000000000010000001100010
?
?
0 1100000000000000000000000000000000000000000011
?
Predicted ????????????????????????????????????????????????????????????????????????????????????????????????????
Prob:
0
.1
.2
.3
.4
.5
.6
.7
.8
.9
1
Group: 0000000000000000000000000000000000000000000000000011111111111111111111111111111111111111111111111111
Predicted Probability is of Membership for 1.00
The Cut Value is .50
Symbols: 0 - .00
1 - 1.00
Each Symbol Represents 1 Case.
Block 3: Method = Enter
Omnibus Tests of Model Coefficients
Chi-square
Step 1
Step
df
Sig.
8.572
4
.073
Block
8.572
4
.073
Model
21.846
9
.009
Page 8
16
Model Summary
Step
1
-2 Log
likelihood
Cox & Snell R
Square
Nagelkerke R
Square
.077
.102
356.815 a
a. Estimation terminated at iteration number 4 because parameter estimates
changed by less than .001.
Hosmer and Lemeshow Test
Step
1
Chi-square
df
2.659
Sig.
8
.954
Contingency Table for Hosmer and Lemeshow Test
being an entertainer = .00
Observed
Step 1
Expected
being an entertainer = 1.00
Observed
Expected
Total
1
21
20.625
6
6.375
27
2
20
18.701
7
8.299
27
3
16
17.202
11
9.798
27
4
17
15.976
10
11.024
27
5
15
14.987
12
12.013
27
6
13
13.892
14
13.108
27
7
11
12.871
16
14.129
27
8
11
11.966
16
15.034
27
9
10
10.492
17
16.508
27
10
12
9.290
19
21.710
31
Classification Tablea
Predicted
being an entertainer
Step 1
Observed
being an entertainer
0
Percentage
Correct
1
0
101
45
69.2
1
57
71
55.5
Overall Percentage
62.8
a. The cut value is .500
Variables in the Equation
95.0% C.I.for EXP(B)
B
Step
1a
newvar
S.E.
Wald
df
Sig.
Exp(B)
Lower
Upper
-.620
.258
5.785
1
.016
.538
.325
.892
e2
.013
.029
.203
1
.653
1.013
.957
1.072
c8
.015
.054
.080
1
.777
1.015
.913
1.130
c12
.057
.045
1.606
1
.205
1.059
.969
1.156
a. Variable(s) entered on step 1: d6, d16, d19, d38.
Page 9
17
Variables in the Equation
95.0% C.I.for EXP(B)
B
Step
1a
S.E.
Wald
df
Sig.
Exp(B)
Lower
Upper
c13
.086
.050
2.944
1
.086
1.090
.988
1.203
d6
.182
.162
1.268
1
.260
1.200
.874
1.647
d16
.276
.133
4.288
1
.038
1.318
1.015
1.711
d19
.075
.133
.318
1
.573
1.078
.830
1.399
d38
-.003
.145
.000
1
.983
.997
.750
1.326
-2.946
1.039
8.044
1
.005
.053
Constant
a. Variable(s) entered on step 1: d6, d16, d19, d38.
Correlation Matrix
Constant
Step 1
newvar
e2
c8
-.495
-.111
1.000
-.118
-.118
1.000
-.111
-.015
-.078
-.377
.008
-.031
c13
-.221
-.094
d6
-.274
-.005
d16
-.273
.104
d19
-.075
.011
d38
-.364
-.079
Constant
1.000
-.014
newvar
-.014
e2
-.495
c8
c12
c12
c13
d6
d16
-.377
-.221
-.274
-.273
-.015
.008
-.094
-.005
.104
-.078
-.031
-.026
-.110
-.012
1.000
-.111
-.086
-.097
.000
-.111
1.000
.037
.053
.103
-.026
-.086
.037
1.000
-.137
.048
-.110
-.097
.053
-.137
1.000
.012
-.012
.000
.103
.048
.012
1.000
-.056
-.066
.074
-.060
-.201
-.290
.120
-.079
.151
-.029
-.145
-.086
Correlation Matrix
d19
Step 1
Constant
newvar
d38
-.075
-.364
.011
-.079
e2
-.056
.120
c8
-.066
-.079
c12
.074
.151
c13
-.060
-.029
d6
-.201
-.145
d16
-.290
-.086
d19
1.000
-.188
d38
-.188
1.000
Page 10
18
Step number: 1
Step number: 1
Observed Groups and Predicted Probabilities
16 ?
?
?
?
?
?
F
?
?
R
12 ?
?
E
?
?
Q
?
11 1
1
?
U
?
1
1 11 1
1 1
?
E
8 ?
0
1 11 1
1 1 11
1
?
N
?
1 0
1 11 1 111 11111
1
?
C
?
1 0
1 11 1 001 1 111 11111
1
?
Y
?
10 0 11110 0 00110 111111011
11
?
4 ?
1 100101 0110000100100 1001110111 11111 1
1
?
?
010000101 01100001000001000001001111110101
1 1
1
?
?
0
00000000000000000010000010000000010110000011111 0
11
?
?
1 0
00100000000000000000000000000000000000000000000000011010 101 1
?
Predicted ????????????????????????????????????????????????????????????????????????????????????????????????????
Prob:
0
.1
.2
.3
.4
.5
.6
.7
.8
.9
1
Group: 0000000000000000000000000000000000000000000000000011111111111111111111111111111111111111111111111111
Predicted Probability is of Membership for 1.00
The Cut Value is .50
Symbols: 0 - .00
1 - 1.00
Each Symbol Represents 1 Case.
Page 11
19
Table 1: Prediction of Being An Entertainer
MEASURES OF MODEL FIT
Block
r
Variable
Exp
(B)
(>1=+)
(<1=-)
Variable
Exp (B)
FINAL
(>1=+)
(<1=-)
BLOCK 1
E1. Gender
(F)
E2. Age
-0.03
.520**
.538*
0.01
1.027
1.013
BLOCK 2
C8.
Spontaneity
.207**
1.035
1.015
C12. Moody
-.178**
1.032
1.059
C13. Social
0.083
1.106*
1.090
BLOCK 3
D6. Gives
Feedback
.231**
1.200
1.200
D16.
Pressure
.323**
1.318*
1.318*
D19. Talks
1-to-1
.264**
1.078
1.078
D38.
Friendly
0.078
.997
.997
Block
ChiSquare
Model
ChiSquare
Model
-2 Log
Likelihood
(lower)
Model
Cox & Snell
R-Square
(higher)
0-near1
(1=perfect)
Model
Nagelkerke
R-Square
(higher)
0-1
(1=perfect)
Model
Hosmer &
Lemeshow
Test
ChiSquare
7.528*
7.528*
371.133*
.027
.036
3.120
5.746
13.274*
365.387*
.047
.063
6.892
8.572
21.846**
356.815**
.077
.102
2.659
incremental
cumulative
cumulative
cumulative
cumulative
cumulative
Key: Significance
*p <.05
**p <.01
NOTE: Hosmer & Lemeshow Test Chi-Square is not significant; meaning the observed and the expected do not
differ significantly (they are similar).
20
Table 2. Classification Result for Being An Entertainer
Predicted
Observed
Actual
Being an entertainer
Group
not being an
entertainer
Overall Percentage
a. The cut value is .500
Being an
not being an
Percentage
entertainer
entertainer
Correct
101
45
69.18%
57
71
55.50%
62.80%
Equation 1. Press’ Q for Logistic Regression
N = Total sample size =274
n = Number of observed correctly classified = 101+71 =172
K= Number of groups = 2
Press’ Q
= [N - (nK)]2/N(K-1)
= [274 - (172*2)]² / (274*(2-1)) = 17.883
df = 1 and 1% Significant Level
Χ2 -critical = 6.63
Press’ Q=17.883 > Χ2 -critical = 6.63
The value of Press’ Q is statistically significant at the 1% significant level. Therefore, we can conclude
the classification is statistically significantly beyond chance.
21
Results Summary: A logistic regression was applied to 274 observations to predict the probability of being an entertainer from variables dealing with gender, age, personality traits, and communication styles. Three blocks of independent variables were utilized. Block one consisted of gender and age. Gender was recoded so that female was 1 and male was 0. Block two consisted of the personality traits spontaneity, moodiness, and social ability. Block three consisted of giving feedback within a conversation, being a relaxed speaker while under pressure, easily communicating in a dyad, and being a friendly communicator. The binary dependent variable I have considered being an entertainer was recoded into a dummy with 1 as I have considered being an entertainer (true) and 0 as I have not considered being an entertainer (false). To test for multicollinearity among independent variables, we ran a separate multiple regression analysis. The results for Tolerance and VIF for all individual independent variables indicate there is no problem with multicollinearity. Block One (consisting of gender and age) is significant at the p<.05 level, with a Block One Chi‐Square value of 7.528. Model One is also significant at the p<.05 level, with a Model One Chi‐Square value of 7.528. The Model One ‐2 Log Likelihood [‐2LL] measure is significant at the p<.05 level with a value of 371.133, indicating that Model One (consisting of only Block One) is a significant model estimation fit. Specifically, Cox & Snell R‐Square (.027) and Nagelkerke R‐Square (.036) estimate that 2.7% to 3.6 % of the variance in the dependent variable is accounted for by Model One. The Model One Hosmer & Lemeshow Chi‐Square Test (3.120) is not significant ‐‐the observed and the expected do not differ significantly‐‐ indicating another measure of good model fit. Gender is a significant and unique contributor to the dependent variable. The gender Exp(B) of .520 is significant at the p<.01 level, and the gender Exp(B) FINAL of .538 is significant at the p<.05 level, both indicating a negative relationship between gender and being an entertainer ‐ for each unit increase in gender (male = 0, female = 1) there is a decrease of 48% percent in the odds of being an entertainer. The formula for calculating the Percentage Change in Odds = (Exp(B) – 1.0) x 100%. Therefore we calculated (.520 – 1.0)x100% = ‐48%. Being female is associated with a 48% decrease in the odds of being an entertainer. Statistically, a male is more likely to have considered being an entertainer. Block Two (consisting of Spontaneity, Moodiness, and Being Social) is not significant, with a Block Two Chi‐Square value of 5.746, suggesting that personality traits alone do not predict the likelihood of being an entertainer. The Exp (B) and Exp (B) FINAL values for the variables Spontaneity, Moodiness, and Being Social are also not significant. However, Model Two (consisting of both Block One and Block Two) is significant at the p<.05 level, with a Model Two Chi‐Square value of 13.274 (testing the Model Two ‐2LL measure of model fit, with a value of 22
365.387, and an improvement over Model One in model estimation fit, since the value of Model Two ‐2LL is less than the value of Model One ‐2LL). The Cox & Snell R‐Square (.047) and Nagelkerke R‐Square (.063) indicate 4.7% to 6.3 % of the dependent variable variance accounted for by Model Two. The Model Two Hosmer & Lemeshow Chi‐Square Test (6.892) is not significant‐‐the observed and the expected do not differ significantly‐‐ indicating another measure of good model fit. Block Three (consisting of giving feedback, relaxed speaker under pressure, dyad, and friendliness) is not significant, with a Block Three Chi‐Square value of 8.572. However, Model Three is significant at the p<.01 level, with a Model Three Chi‐Square value of 21.846, testing the Model Three ‐2LL measure of 356.815, indicating that Model Three (consisting of Block One, Block Two, and Block Three) is a highly significant model estimation fit. Specifically, Cox & Snell R‐Square (.077) and Nagelkerke R‐Square (.102) show a 7.7% to 10.2 % variance accounted for by Model Three. The Model Three Hosmer & Lemeshow Chi‐Square Test (2.659) is not significant ‐‐
the observed and the expected do not differ significantly‐‐ indicating another good measure of model fit. Being a relaxed speaker under pressure is another significant and unique contributor to the dependent variable. The Exp(B) FINAL of 1.318 is significant at the p<.01 level, indicating a positive relationship between gender and being an entertainer ‐‐ for each unit increase (on the likert scale) in being a relaxed speaker under pressure, there is a 31% increase in the odds of being an entertainer. The more a person is a relaxed speaker under pressure, the more likely that person will be an entertainer. The Press’ Q was calculated and valued at 17.883. It is statistically significant at the p<.01 level. Therefore we can conclude that our ability to predict the likelihood of being an entertainer based on these three blocks is statistically significant beyond chance. The findings from this logistic regression analysis are that being both a male, and someone who is a relaxed speaker under pressure, are strong predictors for a person to be an entertainer. And, overall, a model consisting of demographics, personality traits, and communication styles can predict whether a person indicates that they have considered being an entertainer.