Week 3:Probability

Question 1

define inferential statistics

Answer 1

uses a random sample of data (our participants) from a population to help us make inferences about a population to help us reach conclusions that reach beyond our data.

Question 2

define frequentist statistics/what do frequentists do?

Answer 2

assign probabilities to datasets (probability of recurrence in a sequence of experiments)

Question 3

define null hypothesis significance testing

Answer 3

-a method of statistical inferences
-it turns a research question into two hypotheses:alternative and null hypothesis

Question 4

define what a null hypothesis is

Answer 4

a statement of no difference (considered to be true until evidence against it)

Question 5

define what an alternative hypothesis/experimental is

Answer 5

a statement of ‘difference’ ‘association’ or ‘treatment effect’ (it’s what we’re attempting to find evidence for)
you would only ever report this hypothesis, not the null hypothesis

Question 6

define what a one-tailed hypothesis is

Answer 6

states the ‘direction’ the effect/difference/association will be in (“dogs are more likely to live longer than cats”) used for directional hypothesis 5% sig. level conc. on one tail/end

Question 7

define what a two-tailed hypothesis is

Answer 7

doesn’t state a direction (“there’s a difference in the life period of dogs and cats”) used for non-directional (sometimes used regardless of hypothesis) hypothesis 5% sig level split on both tails/ends

Question 8

what type of frequency is collected by nominal data?

Answer 8

event frequencies e.g. frequencies of males/females in a room, smokers/non-smokers etc.
-descriptive statistics presented as frequencies/% of the total

Question 9

why are frequencies used as descriptive?

Answer 9

no other descriptive statistic makes sense

Question 10

what is chi-square?

Answer 10

an inferential statistical test that’s used for nominal (categorical) data and assesses if there’s an association between the categorical variables

Question 11

how to calculate estimated frequencies

Answer 11

(row total) x (column total)/N

Question 12

what does a chi-squared test do?

Answer 12

-measures differences between observed and expected frequencies
-if it differs we can suspect an association between the IV and DV AND between the two categorical variables

Question 13

Explain how there are different types of chi-squares

Answer 13

named by the number of ‘rows’ and the ‘number of columns’.
-e.g., two ‘rows’ (subscribes/does not subscribe) and two ‘columns’ (chocolate/apple)=2x2 chi square
-extra choice/column e.g., apple =3x2 chi square

Question 14

Explain what assumptions are in stats tests

Answer 14

Every stats test has ‘assumptions-rules that our data must
adhere to in order to be analysed using that test so is about looking at your data and making decisions based on certain criteria.

Question 15

State the assumptions of a chi-square test

Answer 15

1. INDEPENDENCE- all frequencies must be unique so can’t be used for within-subjects/repeated measures designs. (as association BETWEEN)
2. All Expected values should be above 5.
   -lots of conditions=all expected values should be greater/equal to 1 AND no more than 20% of expected counts should be less than 5.
   -we need to have an appropriate number of data points but if the sample is too large this can lead to misleading results.

Question 16

define P values

Answer 16

evaluates how well the data from the sample supports the argument that the null hypothesis is true
high p value=data likely supports true null
low p value=data likely rejects true null for the population rather in line with alt. hypothesis (less than 0.05)

Question 17

State the rules of P values

Answer 17

If P value below .05=result is ‘statistically significant’.
-The level at which we accept a result to be significant is known as the alpha level (α)
-however there are current arguments about the usefulness of p values and the problems with them.

Question 18

what does the test statistic let the reader know?

Answer 18

what stats test was conducted (every stats test has one)
e.g., chi square=Chi symbol/X^2

Question 19

why do we use effect sizes?

Answer 19

to show how strong the association is
-2x2 chi square effect size=‘Phi’ φ
-other chi-square designs we can report ‘Cramer’s V’

Question 20

In Cramer’s V, what does the effect size depend on?

Answer 20

the degrees of freedom

Question 21

what’s the main problem with p values?

Answer 21

type 1/2 errors

Question 22

define a type 1 error

Answer 22

the rejection of a true null hypothesis finding (a ‘false positive’)

Question 23

define a type 2 error

Answer 23

failing to reject a null hypothesis (known as a ‘false negative’).

Question 24

how can we reduce a type 1 error occurring? aka Bonferroni correction

Answer 24

change your alpha level accordingly so you no longer accept results to be significant as p<.05.
-divide your alpha level (.05) by the number of tests you are going
to conduct.
■ So imagine I am going to conduct 4 tests on a set of data. I would take the standard alpha level (.05) and divide it by the number of tests I am going to conduct (4).
■ .05/4= .0125 or .013.
■ So I will no longer accept a finding to be significant unless it is
below .013.
■ This is known as a Bonferroni correction.

Question 25

Explain what P hacking is

Answer 25

A method of manipulating data to achieve significant results.
■ Multiple analyses involved
■ Omitting other information (e.g., the existence of other variables).
■ ‘Controlling’ for variables
■ Analyse part way through then gather more data - stop when
you get a significant result
■ Changing the DV (helped via ill-defined / poorly operationalised DV)

Question 26

what’s a solution to p hacking?

Answer 26

could change the alpha level from 0.05 to 0.005 BUT solve with preregistration

Question 27

define preregistration

Answer 27

when researchers predefine what analysis they are going to do before they conduct the study.
■ Outline the method and provide some background research.
■ They should then adhere to this in the paper they publish.

Question 28

what can be the issues with p values?

Answer 28

p values are prone to misuse and questionable research practices.