SCIE5515 L3: Evidence Week 2+3 3/17/26

What are two major developments that reduced age standardized death rate in Australia over the last century?

1) Penicillin

2) Salk vaccine (polio)

Who is this?

-Noticed lower death rate due to childbed fever in midwives' ward compared to doctors' ward

-Cadaverous particles from autopsies spread disease

-Wash hands with chlorine solution in lime water

Ignaz Semmelweis

What hypothesis did Semmelweis form about reducing childbed/puepural fever?

-Wash hands with chlorine/lime solution

-Get rid of cadaverous particles from autopsies

-Reduce death rate from childbed fever

Who is this?

-Pioneered antiseptic surgery

-Introduced carbolic acid

-Reduced sepsis incidence rate

Joseph Lister

Who helped generate evidence for germ theory?

1) Ignaz Semmelweis

2) Joseph Lister

What is this?

-Miasma theory

-Cholera/bubonic plague caused by polluted air

-Diseases common in crowded cities

Name an example of correlation not causation

-Fever trees in African low-veldt

-Malaria is common around fever trees

-Disease actually caused by Anopheles mosquito living in trees

Who is this?

-Used microscopy to provide evidence of germ theory

-Create dot map and statistics of cholera epidemic

-Ended epidemic by removing sewage-polluted water pump

John Snow

Who is this?

-Disproved "spontaneous generation" theory

-Proved microorganisms cause puerperal fever

-Reduces bacterial contamination in milk/wine by boiling

Louis Pasteur

Who is this?

-Discovered inoculation with cowpox scabs achieves immunity against smallpox

-First example of vaccine

Edward Jenner

What vaccine is linked to lower risk of dementia?

Shingles vaccine

Who is this?

-Published MMR vaccine/autism study in The Lancet

-Retracted due to undeclared conflicts of interest

-Struck off medical register

Andrew Wakefield

Name an example of a disease that is now uncommon due to widespread vaccination campaigns

Polio

Who is this?

-"Demarcation problem"

-Distinguish science from non-science via falsifiability

-Can prove hypothesis wrong, not true

Karl Popper

What are the steps to science?

1) Observation

2) Pattern

3) Hypothesis

4) Prediction

5) Test (experiment/observation)

6) Analysis and conclusion

What makes a good experiment?

1) Confidently form conclusions

2) Controls

3) Address confounding variables

4) Replication (generalizability)

What level of the NHMRC Evidence Hierarchy is this?

Level I

-Large meta-analysis of many studies in lower categories

-Example: Systematic review of level II studies

-Cochrane collaboration

Name an example of a Level I study

-Large meta-analysis of studies in lower categories

-Systematic review of level II studies

-Cochrane collaboration

What level of the NHMRC Evidence Hierarchy is this?

Level II

-Random and blind allocation of subjects

-Good control for confounding factors

-Example: Randomized controlled trial

Name an example of a Level II study

-Randomized and blind allocation of subjects

-Controlled trial that addresses confounding factors

-Randomized control study

-Prospective cohort study

-Study of test accuracy with an independent, blinded comparison and valid reference standard among CONSECUTIVE persons with a defined clinical presentation

What level of the NHMRC Evidence Hierarchy is this?

Level III-1

-Pseudorandomized controlled trial

-Has control group, but allocation is not random/blinded

-Example: Alternate allocation

Name an example of a Level III-1 study

-Has control group, but allocation is not random/blinded

-Pseudorandomized controlled trial

-Alternate allocation

-Study of test accuracy with an independent, blinded comparison and valid reference standard among NON-CONSECUTIVE persons with a defined clinical presentation

What level of the NHMRC Evidence Hierarchy is this?

Level III-2

-Comparative study with concurrent controls

-Include subjects who were not exposed/treated

-Examples: Non-randomized experimental trial, cohort study, case-control study, interrupted time series WITH control group

Name an example of a Level III-2 study

-A comparative study with concurrent controls

-Include subjects who were not exposed/treated

-Retrospective cohort study

-Comparative study with case controls (cohort study, and non-randomized experimental trial, interrupted time series with control group)

What level of the NHMRC Evidence Hierarchy is this?

Level III-3

-Comparative study without concurrent controls

-Compares two exposures/treatments

-Examples: Historical control study, two or more single arm study, interrupted time series WITHOUT a control group

Name an example of a Level III-3 study

-Comparative study without concurrent controls

-Compares two exposures/treatments

-Diagnostic case-control study

-Retrospective cohort study

-Case control study

-Comparative study without concurrent controls (historical control study, two or more single arm study, interrupted time series without a parallel control group)

What level of the NHMRC Evidence Hierarchy is this?

Level IV

-Case series with post-test/pre-test outcomes

-Tracks subjects with known exposure, examines medical records for exposure or outcome

-Example: Patients who received similar treatment, case series, cohort study

Name an example of a Level III-4 study

-Case series with post-test/pre-test outcomes

-Tracks subjects with known exposure, examines medical records for exposure or outcome

-Study of diagnostic yield (no reference standard)

-Case series

-Cohort study at different stages of disease

-Cross-sectional study

In a good experimental design, the only difference between groups is (blank)

Manipulated factors (time, season)

What should good experimental design consider?

1) Confounding factors

2) Placebo and blinding

3) Controls

True or False: It is better if the experimenter is unaware of ("blind" to) the treatment group for the experimental subjects

True

What makes statistics inferential?

Infer information from samples about the real world

What do you call an unbiased sample?

Random

What is this?

-Totality of individual observations about which inferences are to be made, existing anywhere

Population

What are some problems with data collection?

1) Bias

2) Sampling errors

3) Statistical errors

It is essential for good experimental design that experimental subjects be sampled at (blank) from the population of interest, and assigned (blank) to treatment groups

Random

What do statistical tests like t-test and ANOVA quantify?

-H0

-Probability of null hypothesis being true

What does p-value from a t-test tell you?

Probability of drawing 2 samples as different as the two samples you are analyzing from a single population

True or False: If there is a high probability of H0 then you conclude that H0 is false

False

True or False: If there is a low probability of H0 then you conclude that H0 is false

True

What does a p value of < 0.05 mean?

5% of the time you will take samples from the same population yet mistakenly conclude that they come from different populations

What is this?

-alpha

-Rate of Type I errors

-Predetermined significance level

What is this?

-Probability of saying two samples come from different populations, when they don't

-Error rate fixed at 5% (0.05)

Type I Error

What describes the rate of occurrence of Type I errors?

Alpha

How do you reduce Type I error rate?

Use lower P value for significance (smaller alpha)

What is this?

-Probability of saying two samples come from the same population, when in fact they don't

-Accept a null hypothesis that is actually false

Type II Error

What is this?

-n

-Difference between two populations

-If it is too small, Type II errors increase

True or False: Small n will increase the rate of Type I errors

False

True or False: Small n will increase the rate of Type II errors

True

Why does a small n result in increased Type II error rate?

-Difference is too small to notice

-End up assuming null hypothesis is true

What is this?

-There is no difference

-Reject null hypothesis, assume there is a difference

-Reject correct H0, false positive

Type I Error

What is this?

-There is actually a difference

-Assume null hypothesis of no difference

-Accept incorrect H0, false negative

Type II Error

What is this?

-Beta

-Rate of Type II Errors

-Calculate POWER of test: subtract from 1

How do you calculate the Power of a test?

1) Calculate 1-Beta

2) Beta = rate of Type II errors

What is a field where power analysis is required for publication?

Ecology

What are advantages/disadvantages to small and large sample size?

Small: Not a waste of time and money

Large: Detect reasonable difference, minimize Type II Error

What are 2 main approaches to estimating required sample size for an experiment?

1) Prior experience

2) Mathematical equations/statistical theory

What 4 pieces of info do we need to calculate n (difference between two separate populations)?

d, s squared, alpha, and beta

What is this?

1) d

Minimum treatment difference that we wish to detect

What is this?

2) s squared

Estimated error variance

What is this?

-alpha level

Acceptable probability of making a Type I Error

What is this?

-Beta level

Acceptable probability of making a Type II Error

What test would you use to compare two means?

T-test