STAT1008 Lecture Notes - Lecture 5: Pie Chart, Prostate Cancer, Dependent And Independent Variables
STAT1008 Week 2 Lecture B
● Placebo and blinding:
○ Control groups should be given a placebo, a fake treatment that
resembles the active treatment as much as possible
○ If possible, make randomised experiments double-blinded -> neither
participants or researchers involved should know which treatment the
patients are actually getting
■ E.g. Green Tea and Prostate cancer -> study conducted on 60 men
with PIN lesions, some of which turn into prostate cancer. ½ given
green tea and others = placebo pill. 9 who got placebo = cancer.
Green tea = one
○ Good randomised experiments allow conclusions about causality
● Types of randomised experiments:
○ Randomising cases into different treatment groups is called randomised
comparative experiment
○ Also give each treatment ot each case, and just randomise the order in
which treatments are received (matching pairs experiment).
■ E.g. Giving both groups of people a kindle version and paper
version -> paper being superior
● Why not always randomised?
○ Randomised experiments are ideal, but sometimes not ethical or possible
○ Have to do the best you can with data from observational studies
● Randomised in data collection:
○ Was the sample randomly sampled?
■ If yes, possible to generalise to the population
■ If no, shouldn’t general to the population
○ Was the explanatory variable randomly assigned?
■ If yes. Possible to make conclusions about causality
■ If no, can’t make conclusions about causality
● One Categorical Variable:
○ Frequency table shows the number of cases that fall in each category
○ Proportion is found by total in that category/overall total (p-hat ->
proportion for sample p with the ^ on top)
○ Proportion for population = p
○ All numbers in relative frequency table = 1
○ Barplot = height of bar is the number of cases falling into categories
find more resources at oneclass.com
find more resources at oneclass.com
Document Summary
Control groups should be given a placebo, a fake treatment that resembles the active treatment as much as possible. If possible, make randomised experiments double-blinded -> neither participants or researchers involved should know which treatment the patients are actually getting. Green tea and prostate cancer -> study conducted on 60 men with pin lesions, some of which turn into prostate cancer. Given green tea and others = placebo pill. Good randomised experiments allow conclusions about causality. Randomising cases into different treatment groups is called randomised comparative experiment. Also give each treatment ot each case, and just randomise the order in which treatments are received (matching pairs experiment). Giving both groups of people a kindle version and paper version -> paper being superior. Randomised experiments are ideal, but sometimes not ethical or possible. Have to do the best you can with data from observational studies. If yes, possible to generalise to the population.