Week 3

Question 1

What is an error often found in Neuroscience research?

Answer 1

interaction effects are not statistically analysed and reported; instead, main effects are compared

Question 2

Which neuroscience fields show erroneous analysis of interaction effects most often?

Answer 2

cellular and molecular neuroscience

Question 3

What is the standard error of the mean?

Answer 3

the variance of the means in a population

Question 4

What is a Bayesian prior?

Answer 4

the probability of an event in a given number of trials

Question 5

in which situations are interaction effects erroneously analysed?

Answer 5

1. comparing correlations
2. comparing effect sizes between pre- & post-test
3. comparing effect sizes between experimental & control conditions
4. when correlating behaviour exclusively with one brain area

Question 6

Can we use standard error bars to compare effects?

Answer 6

no - they only assess between-group differences & are not sensitive to repeated measures

Question 7

Which 2 critical aspects of an experiment are not disclosed through hypothesis testing?

Answer 7

the degree of experimental power and the relationship between a set of population parameters (typically population means)

Question 8

What is PPE and why does Loftus (1993) call for its use?

Answer 8

Plot-Plus-Error Bars procedure: figures with sample means and error bars provide all the info and more that hypothesis testing does

Question 9

What is the question that NHST asks?

Answer 9

Given that H0 is true, what is the probability of these (or more extreme) data?

Question 10

What is meant by P(D|Ho)?

Answer 10

the probability that the data (D) could arise if Ho is true

Question 11

What is meant by P(H0|D)?

Answer 11

the probability that Ho is true given the data

Question 12

Which should we be testing, P(H0|D) or P(D|H0)?

Answer 12

P(H0|D) because we want to see if H0 is true, not examine what happens assuming it is true

Question 13

Why is it difficult to test P(H0|D)?

Answer 13

Because we don’t know P(H0), the probability of H0 before the experiment

Question 14

How can we estimate P(H0)?

Answer 14

through Bayesian statistics

P(H0) is called the prior probability/ the Bayesian prior

Question 15

How are correlations and standardised effect sizes flawed?

Answer 15

they dependent on population variability of the DV

Question 16

How does Cohen (1990) propose we should report effect sizes?

Answer 16

in the form of confidence intervals

Question 17

In what ways does Cohen (1990) believe that confidence intervals are more meaningful than significance tests?

Answer 17

1. they provide information on the null and alternative hypotheses
2. their width is analogous to power analyses: larger sample = smaller CI = larger power

Question 18

What is the issue with testing multiple IVs and DVs?

Answer 18

it increases the likelihood of a type I error

Question 19

What simplifications does Cohen (1990) recommend for all authors?

Answer 19

1. use no more than 2-3 decimals
2. use graphs wherever possible
3. dont compare many variables

Question 20

What is the “Fisherian way” of conducting science?

Answer 20

science proceeds only through inductive inference which is mainly achieved by rejecting null hypotheses, usually at the .05 level

Question 21

What is a misconception about statistical significance?

Answer 21

that the smaller the p-value the more important the finding

Question 22

What does Cohen (1990) recommend when analysing data?

Answer 22

1. estimate your desired population effect size, alpha level, and power level so you can achieve the right sample size
2. effect sizes are more important to report than p-values