Experimental Design

Question 1

Define bias

Answer 1

Any deviation or error from the value which you expect

Question 2

Define experimental error

Answer 2

Arises due to factors other than those you are interested in

Question 3

Define factors of interest

Answer 3

An independent variable that you are actively interested in

Question 4

Define confounding factors

Answer 4

A variable that affects the current outcomes, but is not relevant to the current investigation

Question 5

Define randomisation

Answer 5

A method to reduce bias of the effect of unknown confounding factors. It is a method to randomly select a sample of items of experiments

Question 6

Define replication

Answer 6

The repetition of an experiment under apparently identical circumstances

Question 7

Define blocking

Answer 7

Sometimes it is possible to design the experiment so that all other factors are kept constant and the only thing changing is the factor of interest which is called ‘blocking’

Question 8

Define paired comparison

Answer 8

When the experiments are blocked into twos

This reduces experimental error because it reduces the variation between items/people

Question 9

Define control groups

Answer 9

A group where you are not applying anything new

This is to set a benchmark to assess true differences between the efficiency of the drug

Question 10

Define experimental group

Answer 10

The opposite of a control group
They receive the treatment

Question 11

Define blind trials

Answer 11

Trails where the people in the study are not aware whether they are in the control group or experimental group.

This reduces bias arising as a result of the expectations of the patient

Question 12

Define double blind trials

Answer 12

Trials where neither the patient or the researchers are aware who was in the control group and who was in the experimental group

This reduces bias arising from the expectations of the patients and the administers of the treatments

Question 13

Define completely randomised design

Answer 13

If there are multiple groups/samples, each person/sample element is assigned to a group completely at random. This means that each group is of a different size

This reduces bias

Question 14

Define randomised block design

Answer 14

People/experiments within pairs or groups are kept together but the way that the experiments are allocated within the block are randomised

This reduces bias and experimental error

Question 15

Define effect size

Answer 15

This is a measure of practical significance which is not affected by sample size

The greater the effect size the higher the number

Question 16

Define cohens d

Answer 16

Measure of effect on the difference between two means

Question 17

How to work out cohens d

Answer 17

Two equations in page 7 to find d

Question 18

If d is

0-0.2
0.2-0.5
0.5-0.8
0.8+
What are they called

Answer 18

No effect
Small effect
Medium effect
Large effect

Question 19

When the test/ experiment is statistically significant and practically significant what is the p and d like and what would the conclusion be

Answer 19

Low p and high d

Evidence to suggest a difference in means and the difference may be large/meaningful

Question 20

When the test/ experiment is not statistically significant but practically significant what is the p and d like and what would the conclusion be

Answer 20

High p but low d

Insufficient evidence to suggest a difference but if there were one it would be small. It may be that sample is too small to detect any difference. Further investigation may be valid

Question 21

When the test/ experiment is statistically significant but not practically significant what is the p and d like and what would the conclusion be

Answer 21

Low p and low d

Evidence to suggest a difference in mean but that may not mean anything practically

Question 22

When the test/ experiment is not statistically significant and not practically significant what is the p and d like and what would the conclusion be

Answer 22

High p and low d

Insufficient evidence to suggest a difference in means and if there were one, it wouldn’t be meaningful