Unit 3 (b) - experiments, etc

Question 1

What is an observational study

2 types
explain both

issue

Answer 1

Don’t issue any treatments to units
Can’t infer a causal relationship

Prospective - Follow a sample of individuals into the future collecting data
Retrospective - examine current or past data.

Issue - possibility for confounding. Can’t control those variables.

Question 2

What is an experiment

Answer 2

Treatments are intentionally imposed to subjects that allows to determine a causal relationship between the 2 variables.

Question 3

Who is an exp unit

Answer 3

The individuals/particpants/subjects

Question 4

What is a confounding variable

What does it do

Answer 4

Another variable that is related to the explanatory variable and interferes with the responsible variable.

Creates a false perception of association between explanatory and resp vars.

Question 5

What makes a good experiement

Answer 5

Compares at least 2 treatments - 1 could be the control group (which doesn’t receive any treatment)

Random assignment of treatments

Replication - need to be enough exp units in each treatment group

Control of potential confounding vars (as much as poss)

Question 6

How to randomly allo treatments

Answer 6

Numbers 1 to N
First rand gen a number without replacement- assigned to the first grp
Continue until no more numbers are left

Question 7

What are the adv of randomising

Answer 7

Balances out confounding vars

Question 8

What is the rando block design

Answer 8

Units within each block are similar with regards to a blocking variable

Then random allocation of treatments performed within each block

Question 9

What is placebo

what does it allow us to do

Answer 9

Fake treatment that is similar to treatment being used

Helps determine if an effect is truly due to the treatment not simply because of the placebo effect

Question 10

What is a single blind exp

What is a double blind exp

Answer 10

Single blind - Units are not aware of the treatment they are receiving but experimenters are (or vice versa)

Double blind - Neither subj nor researchers know who the subj are, or what treatment is being administered to who.

Question 11

What is the matched pair exp

2 types

Answer 11

Find 2 blocks that are very similar to each other
Randomly assign treatments within each block
Total of 2 treatments

OR

Giving each exp unit both treatments in a random order.

Question 12

What is statistical significance

Answer 12

Diff between exp treatment groups are evidence that treatment causes effect

Question 13

What does rando assignment allow us to do

Answer 13

Allows us to conclude that large observed changes are not merely by chance.

Question 14

What does a rando block design allow us to do

Answer 14

Helps to separate natural variability from diff due to the blocking var.