Binary Logisitic Regression on a Categorical Variable with 3 Values
(Pass/Fail on x = 'Sex Move')
Binary Logistic Regression
What we have looked at thus far in this exploratory analysis were 2 × 2 tables. Now we are going to move to 2 × 3 tables.
First we will tally the discrete variable Moving. Moving was coded as 1 if the person said that the glass was not moving when they drew the line, and 2 if it was. 29 out of the 166 subjects said that the glass was moving.
Moving
|
Count |
1
|
137
|
2
|
29
|
N = 166
|
Now we will create a new variable called 'sexMove' as follows: Gender is coded 0 = female and 1 = male. Moving was coded as 1 if the person said that the glass was not moving when they drew the line, and 2 if it was. We will let the combined 'Gender by Move' = 10*Gender + Move.
According to the dataset 79 females said the glass was not moving, 28 females said the glass was moving. 58 of the males said the glass was not moving and only 1 male said the glass was moving
Female, Not moving
|
79 |
Female, Moving
|
28
|
Male, Not moving`
|
58
|
Male, Moving
|
1
|
N = 166
|
For the purposes of this analysis we will combine the last two rows and label it Male such that this new variable, SexMove, will have 3 values, 1, 2 and 3.
Value
|
Description
|
Count
|
1
|
if the person is female and said the glass was not moving |
79
|
2
|
if the person is female and said the glass was moving |
28
|
3
|
if the person is male |
59
|
We can run the binary logisitic regression using the SAS program ???
SAS program image here...
SAS output and discussion here ...
Conclusion
There is a very highly significant difference in the proportions of persons passing for the three values of SexMove. Only 7.14% of females who said the glass was moving passed the water level task. 37.97% of the females who said the glass was not moving passed, and 64.41% of the males passed the task. Only one male out of 59 said the glass was moving compared to 28 out of 107 females.