weeks 4,5,6 - experimental design

Question 1

definition of a hypothesis

Answer 1

clearly stated explanation, based on observations and assumptions, that leads to a testable prediction.

Question 2

null hypothesis

Answer 2

thought as the hypothesis stating that nothing is going on.

Question 3

scientific method

Answer 3

1. observation
2. question
3. hypothesis
4. prediction
5. test
6. results

Question 4

what three structures need to be thought carefully about when designing an experiment?

Answer 4

1. treatment structure = what treatments will be used?
2. design structure = how do we decide which treatment each experimental unit will get?
3. response structure = how do we measure the response?

Question 5

indirect measure

Answer 5

a measure taken on a variable which can be used as an indicator of the state of another variable that is difficult or impossible to measure

Question 6

manipulative experiments

Answer 6

artificially change something about the experimental units and study the effect of this change.

Question 7

correlative experiments (observational)

Answer 7

use naturally occurring variation, rather than artificially creating variation

Question 8

advantages and disadvantages of manipulative studies

Answer 8

ADVANTAGES
- exclude confounding factors
- effect size can be influenced by the level of treatment
DISADVANTAGES
- doesn’t necessarily display biological variation
- needs good controls
- can bring ethical issues

Question 9

advantages and disadvantages of observational studies.

Answer 9

ADVANTAGES
- easier to carry out (less work, time, effort)
- less time handling animals/organisms
- deals with biologically relevant variation

DISADVANTAGES
- reverse causation effects can occur
- can suffer from third variables/confounding factors

Question 10

pilot study

Answer 10

• allows you to become familiar with the study.
• aid in experimental design.
• allows you to fine tune data collection techniques.

Question 11

experimental units

Answer 11

the units which are assigned different treatments, and whose responses are then measured and compared using statistical methods.

Question 12

what are independent data points?

Answer 12

• a key assumption of statistical tests.
• come from unconnected experimental units

Question 13

pseudo replication

Answer 13

non-independent data points violate statistical test assumptions.

Question 14

random variation

Answer 14

quantifies the extent to which individual subjects in our sample differ from each other for reasons other than the one we are interested in.

Question 15

replication and why we need it

Answer 15

repetition of experimental units within a single treatment.

• allows us to estimate experimental error and increases precision.
• captures natural variability
• protects experiment from chance events

Question 16

randomisation and how it helps

Answer 16

random selection of study sites and individuals.

• helps to achieve independent observations.
• reduce bias and increases accuracy.

Question 17

confounding factors

Answer 17

a factor which is not the one of interest which also has a impact on the factor we a re measuring.

• make it difficult for us to interpret our results.

Question 18

interspersion

Answer 18

often some underlying spatial heterogeneity (variation) in our environment.

to reduce impact on the environment, replicates of different treatments are interspersed.

Question 19

three kinds of manipulative studies when analysing resource availability impact on leaf shape.

Answer 19

1. lab experiments
2. greenhouse experiments
3. field experiments

Question 20

three kinds of correlational studies when analysing resource availability impact on leaf shape.

Answer 20

1. natural experiments
2. observational studies

Question 21

advantages and disadvantages of lab experiments.

Answer 21

ADVANTAGES
- maximum control of conditions
DISADVANTAGES
- expensive
- relatively small sample size

Question 22

advantages and disadvantages of greenhouse experiments

Answer 22

ADVANTAGES
- cheap
- larger sample size
- often similar explanatory power as lab experiments

DISADVANTAGES
- still expensive
- high maintenance
- limited space

Question 23

advantages and disadvantages of field experiments

Answer 23

ADVANTAGES
- more realistic

DISADVANTAGES
- relatively little control (multifactorial and interconnected confounding factors)

Question 24

advantages and disadvantages of natural experiments

Answer 24

ADVANTAGES
- comparison of systematically varying natural conditions
- measurable variation despite low degree of experimental control
- natural situation reflects natural complexity

DISADVANTAGES
- difficult to determine the influence of confounding factors

Question 25

advantages and disadvantages of observational studies

Answer 25

ADVANTAGES - measurable variation - natural situation reflects natural complexity DISADVANTAGES - difficult to determine the influence of confounding factors - no experimental control

Question 26

what kind of study offers the most control?

Answer 26

laboratory manipulative

Question 27

what kind of study captures the most amount of natural complexity of a system?

Answer 27

observational studies

Question 28

as replication 1........ variation 2........

Answer 28

1. increases 2. decreases

Question 29

effect size

Answer 29

magnitude of the effect we are measuring.

Question 30

standard deviation

Answer 30

spread of values around the mean.

Question 31

what happens to the standard deviation when replication is increased

Answer 31

standard deviation gets smaller.

Question 32

what happens to estimation of treatment effects when sample size increases?

Answer 32

estimation of treatment effects become more precise

Question 33

what there things must a sample of a population be?

Answer 33

1. large enough - higher the standard deviation, the larger the sample size must be. 2. representative - random choice of samples bias 3. independent - value of one replicate does not change your expectation of the other

Question 34

simple random selection

Answer 34

all individuals of a population identified and fixed number of samples are randomly selected.

Question 35

stratified sampling

Answer 35

separate population dependent on variable of interest and then select at random within each strata which have been created.

Question 36

inferential statistics

Answer 36

use measurements from a subset of subjects to make generalisation about larger populations of subjects.

Question 37

pseudoreplication

Answer 37

use of inferential statistics to test for treatment effects with data from experiments where either the treatments are not replicated or replicates are not statistically independent.

Question 38

what are the different kinds of pseudoreplication?

Answer 38

1. simple - one unit of replication between a control and a treatment. 2. sacrificial two replicates per treatment and control and two samples within each replicate. 3. temporal - several replicates are taken of the same sampling unit over time and these measurements over time are used pseudoreplicates.

Question 39

what is statistical power?

Answer 39

the probability that particular experiment will detect an effect. OR it is the probability that we will correctly reject the null hypothesis.

Question 40

if you found the expected results 72 out of 100 times, what would the statistical power be?

Answer 40

0.72 = number of expected results/total number of times we have sampled.

Question 41

type I error = false positive

Answer 41

rejecting the null hypothesis when its actually true.

Question 42

type II error = false negative

Answer 42

not rejecting the null hypothesis when its actually false.

Question 43

what three factors can impact the power of an experiment?

Answer 43

1. effect size 2. amount of random variation 3. number of replicates

Question 44

how does effect size effect the power of an experiment?

Answer 44

- larger effect size leads to easier detection of changes. - power is the probability that we will correctly reject the null hypothesis. - high probability of correctly rejecting the null hypothesis, therefore means we have a large amount of power.

Question 45

how does random variation affect the statistical power?

Answer 45

as random variation increases, detecting an effect becomes more difficult, therefore statistical power is lowered.

Question 46

how can statistical power be increased?

Answer 46

- increasing number of measurements (replicates) we collect - reduce random variation

Question 47

how can we reduce random variation?

Answer 47

- pilot study - calibration - standardise conditions across your study (laboratories)

Question 48

what statistical method would a two factor designed experiment use?

Answer 48

two way ANOVA