Logistic Regression

Question 1

Contingency table assume independence of _______and _________ and compare __________ and __________ values to calculate __________ (for counts use ____ correction for continuity)

Answer 1

Contingency table assumes rows and columns are independent and compare observed and expected value to calculate chi^2
for counts use Yate’s correction for continuity

Question 2

How is significance of odds ratio indicated using chi squared

Answer 2

1. Chi^2 uses ‘expected values’ calculated assuming the two odds (A/C and B/D) are the same.
   If odds were the same, the ratio =1
2. chi squared assess the probability that odd ratio is one.
   if p<0.05 then there is a significant change in the odds

Question 3

How to calculate relative risk

Answer 3

RR = p(lung cancer)smokers/ p(lung cancer)nonsmokers

Question 4

How to calculate odds

Answer 4

with a binary outcome variable there are two and only two possible outcomes
each outcome has a probability
the outcomes are mutually exclusive
odds = p/(1-p)

Question 5

Relative risk indicates ________________, and is more ______ than odds ratio. We cannot easily calculate the _____ of RR

Answer 5

Relative risk indicates the probability of something happening (chances of dying is 6.2 times greater if you smoke than not),
relative risk is more intuitive than odds ratio,
however, can’t easily calculate the significance of Relative risk

Question 6

what is odd ratio

Answer 6

express effects as changes in the odds. the chances that smokers die by age 60 are 31.6 times greater than non-smokers

Question 7

Logistic regression use _______ DV and _________IV

Answer 7

binary DV and continuous IV

Question 8

Why cant we use normal regression for binary DV and continuous IV

Answer 8

can get probability 1, can get impossible data points

on a technical note, expect the residuals to depend on the X value, therefore, cant trust the result

Question 9

why is logistic regression good for binary DV and continuous IV

Answer 9

value restricted in the range (0..1)
input can take random values
use the logistic regression equation p(event)=1/(1+e^(-z))

Question 10

Describe the relation between logistic regression and odds

Answer 10

1. the logit is the logarithim of the odds
   logit(p)=z= ln(p/(1-p)) = a1X1+ a2X2 +… +b
   in other words, logistic regression is estimating the odds as a function of predictors
2. the ratio of the odds per unit increase is constant for all values, because the odds are an exponential (e^z)

Question 11

How to determine significance of odds ratio in logistic regression

Answer 11

1. Because the odds ratio is constant, every estimate is based on the fit of the logistic equation
2. use the significance of the logistic regression to determine if the odds ratio is different from unity: non significant fit implies no change and IV has no effect
3. significance of odd ratio given by significance of regression