P Values and Comparing Groups

Why Compare the Health of Groups?

Compare ethe Health of 2* Groups- One Time-Point per Patient Cross-Sectional Study

Cross-sectional study—>One time point.  one group surveyed to test associations between exposures and outcome/s

• Make 2 measurements per person: amount of cause vs amount of effect

• 2 measurements: predictor vs. outcome

Compare ethe Health of 2* Groups- One Time-Point per Patient- Ecological Study

Ecological study—>community/population observed to test associations between exposures and outcome/s

Compare ethe Health of 2* Groups- One Time-Point per Patient

1. Both studies focus on simultaneous observation of exposure and outcome, but difference is unit of observation

• Cross-sectional study – focus on individual/patient level data

• Ecological study – focus on population-level data (e.g. by country)

Compare ethe Health of 2* Groups- Two or More Time-Points per Patient- Cohort Study

Cohort study—>Prospective. 2+ time points.  Start disease-free participants followed up to see if they develop a disease (condition/outcome) of interest

• Usually with groups who differ at outset on some exposure/s of interest

Compare ethe Health of 2* Groups- Two or More Time-Points per Patient- Case-Control Study

Compare ethe Health of 2* Groups- Two or More Time-Points per Patient- Randomised Controlled Trial

Randomised controlled trial (RCT)—>Prospective. 2+ time points. groups who are randomly allocated to receive intervention(s) versus comparator(s)

• Test safety and efficacy/effectiveness of interventions

• Example: Ronald Fisher’s fields to test efficacy of manures

Risk VS Odds- Smoking —> Lung Cancer

What are the Strengths of Case Control Studies?

1. Useful when outcome/disease is rare

2. Can offer some evidence of cause – effect relationship i.e. association between exposure and outcome

3. Can identify multiple exposures (both positive and negative associations)

4. Minimises selection and information bias

5. Retrospective – much cheaper and typically shorter in duration

What are the Weaknesses of Case Control Studies?

1. Less suitable for rare predictors/exposures

2. Can be hard to ensure exposure occurred before onset

3. Cannot calculate prevalence or incidence

4. Vulnerable to confounding

5. Suitable control group may be difficult to find

6. Retrospective data availability and quality may be poor

What is a Randomised Controlled Trial?

Why and When are RCTs Conducted?

What are the Strengths of RCTs?

1. Minimise selection bias

2. Minimise measurement bias

3. Best single-study evidence for causal association between exposure (intervention) and outcome

What are the Weaknesses of RCTs?

1. Time-consuming

2. Difficult

3. Expensive

4. Researchers can over-rely on them

What is the Goal of Statistical Testing?

What is a Variable?

A variable = A parameter that you can or will measure/observe, which varies in your cohort

• Height, plasma glucose, eye colour, alive or dead (“living status”), treated or untreated

What is an Outcome Variable?

An outcome variable is the variable that supposedly results from other inputs.  It occurs later.

• What you think is “the effect that resulted from what you did”

• Also called the dependent variable

What is a Predictor Variable?

A predictor variable is any variable that may control your outcome variable.  It occurs before the outcome

• A “predictor” is what you think might be “the cause”

• You might control it, or cause it to vary, OR

• It might be determined naturally, but you select for it

• Also called independent variable or explanatory variable

What is a Confounding Variable?

Confounding Variable

Another (unwanted & unmeasured) variable or fact that is messing up your ability to correctly reach a conclusion

What is a Continuous Variable?

A continuous variable is a number that can go through many values and fractions: height, age

What is a Categorical Variable?

A categorical variable is a parameter whose values are limited:  eye colour, country of birth

What is Binary or Dichotomous Variables?

Binary or Dichotomous Variables – a parameter with only 2 possible values

• Alive or dead; Are you a parent: yes/no; diagnosis of bipolar disorder (yes/no); tobacco smoker (yes/no); heads or tails

• Dichotomous variables and binary variables are synonyms

• You could say they are both a subset of categorical variables (which often have 3 or more possible values)

What are Ordinal Variables?

Ordinal variables

• A variable where the values have an order but not a clear numerical relationship, e.g. socio-economic status

• These variables function like rank in the military: a colonel is higher than a sergeant, but is it twice as high?

• Likert scales are ordinal: strongly disagree, slightly disagree, neutral, slightly agree, strongly agree

Why are Statistical Tests Used?

What is the Chi Squared Test?

What is the Difference between Type 1 and Type 2 Error?

Errors Based on Your Data- Type I vs Type II

What is the T-Test?

What does the T-Test Compare?

What is a P Value?

P Value- What You Want VS What You Get

Why are Small Values of P Good?

P Values- Example Comparing Two Randomly Generated Groups

What is the Null Hypothesis?

What is the Effect Size?

Effect Size- Cohen’s d

P Values VS Effect Sizes

What are Confidence Intervals (IC)?

How do you Compare Health Outcomes of Groups?

Summary