1003 Stat Flashcards
(92 cards)
1
Q
5 parts of an experimental
A
- Reliable - something that is going to happen again and again (consistently)
- Valid - confident that our results mean what we really mean
- Parsimonious - theory must be as simple as possible while still using good ideas
- Cumulative - builds on prior research and build on prior mistakes
- Public - open to scrutiny of the wider scientific community
2
Q
Reliability
A
refers to our confidence that a given finding can be reproduced again and again
3
Q
Validity
A
refers to our confidence that a given finding shows what we believe it to show
4
Q
Three types of Validity
A
- Internal validity - does the outcome really reflect the experimental manipulation
Can we determine casualty
A more tightly controlled study has more internal validity (every confounding variable is removed and it is only the two variables affecting each other)
Tightly controlled makes a clear validity - External validity - how well can we generalise the findings to other people/situation
- Construct validity - does the theory (construct) relate to the measurement being used
E.g. does an IQ test actually reveal intelligence
5
Q
Designing a study
A
- Identify research question
- Define IV
- Define DV
- Choose a sample
- How will the results be interpreted
6
Q
cumulative research
A
using past research
7
Q
Operationalisation
A
defining how a concept will be measured
8
Q
hypotesis
A
what we expect based on past knowledge
9
Q
Define the IV and DV
A
- what we want to measure
- what we can measure (can it be measured)
- what we should measure (ethics)
10
Q
Between-subjects design
A
There are two different groups of participants in the control and experimental groups, and the experimental manipulation occurs between these groups
11
Q
Within-Subjects design
A
The same group of participants are both the control and experimental groups. The experimental manipulation occurs within the same group
12
Q
Relevance-sensitivity trade-off
A
The more sensitive a DV is to changes in the IV the less relevant it may be to the real-world phenomena
13
Q
Threats to internal validity
A
- Due to time e.g. fatigue effects
- Due to experimental situation e.g. testing effects
- How the sample is chosen e.g. morality
14
Q
Threats to external validity
A
- experimenter bais
- demand characteristics
- interaction
15
Q
Practice effects
A
the more someone does a task the better they get and therefore will perform better next time they are doing the task
16
Q
fatigue effects
A
the person gets tired of doing the task
17
Q
Maturation affects
A
the people in the test mature overtime and therefore are better able to do the task
18
Q
History affects
A
history effects like Covid 19 may change the way the task is done the second time
19
Q
Cognitive perspective
A
Focus is on how people perceive, process, store and retrieve information
20
Q
Types of data
A
- Nominal
- ordinal
- interval
- ratio
21
Q
Nominal
A
- no ordering
- discrete
- e.g. university you attend
22
Q
Ordinal
A
non-consistent ordering
e.g. rank in army
discrete or continuous
23
Q
Interval def
A
- consistent ordering
- e.g. temperature (C)
- discrete or continuous
24
Q
Ratio
A
Consistent ordering
height
true zero
discrete or continuous
25
X def
mean
26
Distribution
Perfect symmetrical data has a single mode/one hump
27
Negative skew
left tail
mean < median < mode
28
Positive skew
right tail
mode < median < mean
29
Variance
the average of the squared distances between each score and the mean
STD squared
30
Standard deviation
is the variance of the dataset
variance square root
31
Descriptive statistics
describe your sample
32
inferential stats
Use your sample to describe the population
tells use whether results are produced by chance
33
Population STD
that o symbol (sigma)
34
How to correct bias
N-1 which helps correct for our incorrect estimation
35
Normal distribution
Use the sample data so we can estimate theoretical distribution to reflect the population distribution/the real world
36
unimodal
one hump
37
symmetrical
normal distribution
mode=mean=median
38
effect of high STD
more spread out the bell curve will be
39
z score
tell you how many standard deviations the score is away from the population mean
formula: sample mean divided by population std (individual)
formula: sample mean divided by population SE (group)
40
Standard error formula
STD divided by the square root of N
41
Steps for Null Hypothesis Significance Testing
1. Define the null hypothesis
2. Define the alternative hypothesis
3. Design study and collect data
4. Calculate t-stat
5. compare this to the critical value
6. Decide if you reject or retain the null hypothesis
42
t-test
A statistical test that compares two means and tells you if the difference is statistically significant
used when the std is unknown
43
null hypothesis written out
u1 = u2 = u
(group 1 population means = groups 2 population means = general populations mean)
44
alternative hypothesis written out
u1 doesnt = U2
45
sample size on the peak of graph
small sample = wider and flatter the graph
46
T-value formula
information term divided by error term
47
Info term
difference between our group means
48
Error term
what could have happened by chance
49
What does it mean when T=0
no difference between groups
50
3 types of t-tests
1. one-sample t-tests
2. dependent/paired samples t-test (within participants)
3. Independent samples t-test (Between participants)
51
one-sample t-tests
For measuring a single group and you are comparing to a known standard
52
Dependent/paired samples t-test (within participants)
For when you are comparing the same group to themselves at two time-points
53
Independent samples t-test (Between participants)
For when you are comparing two different groups to one another
54
Critical T Value
a value that cuts-off some proportion of the distribution that you are comparing your test stat to
t-distribution tells use the proportion of scores that lie beyond the critical value
55
The rejection region
the proportion of our distribution that gets cut off by the critical value
56
one tailed t-test
if the rejection region is on a large part of one side (1 critical value)
57
Two tailed test
if the rejection region is on a even amount of two sides (2 critical values)
58
retain the null when
our observed t-value lies inside the critical t-value
59
reject the null hypothesis when
our observed t value lies outside the critical t-value
60
p-values
The proportion of the t-distribution that the t-test cuts of
61
Type I error
incorrectly reject the null
True
Pregnant man
62
Type II error
Incorrectly accept the null
False
Not pregnant pregnant lady
63
Effect size
a measure of how big the difference is between your two conditions/groups
64
Cohen's D
Expresses the difference between means in terms of STD units
65
Cohen's D ranges
0.2 - small
0.5 medium
0.8 large
66
Confidence Intervals
the probable range of values for a population parameter
67
Width of CI
The larger the level the more confident you are of all the possible you parameter could take
68
Survey
Researcher collects info about different variables from a number of different people to examine the relationship between the variables
69
Difference between survey and experiment
Experiment - measure DV and manipulate IV (emphasis on causation)
Survey - you measure both the IV and DV (emphasis on relationship
70
Challenges for design survey
1. sample Selection
2. Sample size
71
Sample selection
1. systematic probability (simple random)
2. Systematic non-probability (purposive, snowball)
3. Convenient non probability (convenience sampling)
72
Purposive
going out to the population to find exactly who you want to survey
73
Convenience
getting anyone they possibly could get
74
Inference
drawing a conclusion about the population using the survey
75
Non responder problems
1. Mortality - participants drop out for some reason
2. Reactivity - participants have bad behaviour e.g. lazy people
76
3 Survey types
1.Interview - sit down and ask them question
2. Computer-aided interview - they complete it on a computer/phone
3. Naturalistic observation - they carry a microphone for the day
77
overcoming people responding how they think they should rather as they actually would
Deception studies or non-obtrusive measures
- Measures that ensure participants are not aware that they're being involved in research
Archival records
- Studies that use these techniques check through records of behaviour
- E.g. look at past academic results
78
Questions to avoid
Ambiguous questions - vague or imprecise wording
Double-barrelled questions - asking multiple questions in one
Leading questions - questions that bias a response
79
conditions for two variables to be correlated
1. As values of one variable change, second variable values change too
2. Values of one variable (criterion) are predictable based on values of another (predictor)
80
Conducting correlation analysis
- visualise data on scatterplot to asses if variables are linearly related
- Calculate regression to plot the regression line (line of best fit)
- convert scores for both variables into Z scores for pearson's
- asses significance of correlation using NHST framework
- calculate effect size measure (r squared) in order to estimate the proportion of variance in criterion explained by predictor
81
correlation def
the relationship between two variables
82
Scatterplot
a graph that plots each participants score on both variables as a point in two-dimensional space
83
function of scatterplot
o Visually inspect if relationships are linear (which is expected)
o See if there are outliers or problems that could skew calculations
o Indicate strength of relationship
84
Regression equation
Y = a + bX
b - slope of the line
a - y-intercept
85
Pearson's correlation coefficient (r)
- range from -1 to 1
- strength and direction relationship
86
positive correlation
A relationship between variables where high scores on the first variable are
associated with high scores on the second variable, and low scores on the first are associated with low
scores on the second
87
Negative correlation
A relationship between variables where high scores on the first variable are
associated with low scores on the second variable, and low scores on the first are associated with high
scores on the second
88
Calculation for pearson r
ZxZy / N-1
89
Chi-squared tests
Test used to determine if there's a significant association between two categorical variables
90
Chi Squared formula
Xsquared = (O-E)² / E
91
Oberserved frequencies (O)
how many people are observed in each category
92
Expected frequencies
How many people do we expect to be in each category
