Measurements

[Risk and Odds Ratio]

Risk and odds ratio is the measurement of association between two categorical variables
Categorical variables are variables that consist of group or category names

RISK(A|B) = (A and B) / B
RISK (A|C) = (A and C) / C
Risk ratio => RISK(A|B)  / RISK(A|C)

e.g gender
Example: Lung cancer and drinking

Risk Ratio:
Risk of Lung cancer for people who drink=> Chances of getting lung cancer/people who drink
Risk Ratio = > Risk for people who drink / risk of people who dont drink

Risk = 1 means no association

Using risk:
Total num of B * Risk(A|B)  = estimated of B who have A

Cohort and Case-control studies:
Cohort: Samples are from exposure groups
Risk and RR can be estimated from sample table

Case-control: Samples from disease groups
It's good for rare diseases
There are lesser samples this rr is hard to determine


Odds ratio:

Odd(A) among B = (Aand B) / ( not A and B)
Odds ratio = Odd(A) among B / Odd(A) among C

Example:
Odds of getting lung ca for drinkers=> drinkers who have lung cancer/drinker who don't have
Odds ratio => odds of getting lung cancer for drinkers/odds of getting lung cancer for non-drinkers

Risk and Odds:
Odds(A) among B = risk(A|B) / 1 - risk(A|B)

Interpreting OR
= 1 (No diff in disease risk between the two group)
> 1 (Higher risk in first group)
< 1 (Lower risk in the first group)

Cross product ratio:
Odds ratio = (A and B) * (not A and C) / ( not A and B) * (A and C)

An odds ratio is good for case-control studies (rare diseases)

Cross product is very useful in Multi-level contingency table

[Variable types]

1)Categorial
Ordinal: Categories with ranges and level in order
e.g Education level, level of income, happiness level

Nominal:
A fixed type of categories
e.g Sex, Blood type

2)Numerical:
A fixed number
e.g height, BMI

[Sources of variability]

True score = True value + random error + systematic error

1) Random Error

Cased by any factor that randomly affects measurement but can be cancelled out by averaging

2) Systematic error

A factor that systematically lowers the score

3) Natural Variability

An affecting factor that is not something that we can control and occurs naturally

e.g Weight (Not everyone has the same weight)

 Natural Variability can also happen if the variable depends on time

[Biases(pitfalls)]

1) Deliberate Bias

Questions can be deliberated worded to support a certain cause

e.g 

"Do you agree that abortion, the murder of innocent beings, should be outlawed?"

-> Biased as its clearly perceived abortion as a bad things

2) Unintentional Bias

A question that is worded such that the meaning can be misinterpreted by many

e.g

"Do you use drugs?"

-> what type of drugs?

3) Desire to please

Many respondents have a desire to please the person who is asking the question

e,g

People tend to understate responses about undesirabble social habits

" How many days a week do you smoke on estimate?"

4) Asking the uninformed

People do not like to admit they dont know what your talking about

5) Unnecessary complexity

If questions are to be understood, they must be kept simple

e.g

"Shouldn't former drug dealers not be allowed to work in hospitals after they are released from prison?" -> Confusing

6) Order of questions

e.g 

" 1.To what extend do you think teens today worry about peer pressures related to drinking alcohol?"

" 2. Name the tip five pressures you think face teens today"

-> we plant ideas into the respondant's head

7) Confidentiality and anonymity

People answer differently based on the degree to which they are anonymous

Confidentiality - > Researches knows their identity but promises not to release the information to others

Anonymity-> researchers don't know the identity

Topics such as sexual behaviour might be hard to conducts

<Prev Study design                        Next Sampling >