Test 2 Flashcards
Quantitative data
numerical
ex. scales, physiological responses
Qualitative data
any form other than numbers
ex. words, sounds, images
qualitative
1. strengths
2. weaknesses
- rich in detail, captures human experiences, why people do things
- resource intensive, smaller samples (*more generalizability), less agreement on appropriate methods of analyzation
quantitative
1. strengths
2. weaknesses
- can use statistics, how often, how much, more agreement on methods
- numbers provide limited information
Qualitative data can be transformed into numbers
ex. watch a movie, but count number of times things happen
gain use of statistics, but lose potentially important information
Coding schemes
a set of rules for turning qualitative to quantitive (rules on how you measure.
*the specificity reduces the measurement error and improves reliability **you need to be specific
qualitative analysis
*perserving detail
- inductive (be objective)
- deductive (go into it having a theory)
- thematic analysis (read data and identify common themes)
- reflexivity (acknowledging potential bias)
mixed-methods approach
combine qualitative and quantitative
do qualitative + quantitative analysis of qualitative data
ex. get them to rate something, then ask an open-ended question
correlation studies
measure 2 or more variables and how they relate
*naturally occurring, no manipulation
- positive correlation
- negative correlation
- Pearson correlation (r)
- as 1 variable gets larger, so does the other
- 1 variable goes up, the other goes down
- measures the linear relationship between 2 variables
- positive values
- negative values
- values range from 0 to 1
- values range from -1 to 0
- increases in 1 variable means…..
- perfect correlation are…..
- limitation:
- highly predictable second variable
- impossible in the real world - complexity of human behaviour, measurement error
- only detects linear relationships
- correlations do not…..
- correlation of 0
- correlation of 1
- represent percentages
- no relationship
- perfect relationship
- effect size
- Cohen
- Hemphill
- very few stronger than…..
- measuring the strength of the relationship/ association
- small 0.1, medium 0.3, and large 0.5
- bottom third, middle third, top third
- 0.3
square of correlation r2
percentage of variance accounted for. Tells you how much the variance in the dependent variable is accounted for by the independent variable.
For example, a model with an R-squared value of 0.9 means that approximately 90% of the variance in the dependent variable is explained by the independent variables. This suggests a strong relationship between the variables and indicates that the model provides a good fit to the data.
You need _____ in both things you are looking at in order to examine correlation
variability
_____ differences in correlation correspond to _____ differences in variance
small, large
correlation studies limitations
1.
2.
3.
4.
5.
6.
- correlation is not causation
- control for third variables (partial correlation & multiple regression)
- temporal information (something needs to cause A before you can measure B)
- linear assumption
- need variability to see correlation
- outliers
- Curvilinear relationships best seen through….
- important to examine ___ before ____
- scatter plots
- data, analysis
correlation studies strengths
1.
2.
3.
- naturally occurring
- useful when things cannot be manipulated or changed
- can be applied to archival data
- Naturalistic observation
- observations taken over a _____ period are more ____ than a single observation
- observations for descriptive purposes, with interpretations
- longer, accurate
- Observation in the lab =
- Observation outside the lab =
- naturalistic observation
- systematic observation
measure naturalistic observation by:
1.
2.
3.
- coding schemes (quantitative)
- interviews (quantitative)
- no interfering, no influences
N.O
1. strengths
2. weaknesses
- realism/ecological validity, often rich qualitative detail, exempt from ethical review
- difficult, time-consuming, can be lots of data thats hard to interpret
Case studies
extremely detailed description of a person of a few people. Necessary for unique/rare populations. Can be qualitative or quantitative
- nomothetic studies
- reductionism
- studies that are from groups of people, but they are averages of a lot of people - which loses detail
- finding averages - but losing detail
case studies
1. strengths
2. weaknesses
- lots of detail, rare populations, generate & test hypothesis
- generalizibility
- Archival data
- content analysis
- existing data surrounds, usually publicly available
- systematically observing and quantifying (coding schemes), counting instances of interest, analyzing texts (natural language processing)
archival data
1. strengths
2. weaknesses
- ecological validity (real world), lots of data
- difficult to organize and obtain, only have what is available, cannot make casual inferences
survery & questionnaires
most simple way to gather info about a person. *but they must know this info and be willing to share it.
response sets
ways of responding dont reflect the question content
Likert scales need
1.
2.
3.
4.
- neutral mid-point
- number of points depends on question
- label all points
- is behaviour low-frequency or high-frequency
scales of measurement
1.
2.
3.
4.
- nominal = named categories
- ordinal = ranked order
- interval = equally spaced (continuous)
- ratio = interval with a meaningful zero (continuous)
descriptive statistics
summarizing your data - an average across items or across items and participants
includes correlation studies
central tendency
- mean = average (sum / # of scores)
- median = middle *helpful with extreme scores
- mode = most frequent score
- variability
- standard deviation
- low standard deviation means..
- how scores differ from one another, spread of scores, difference in responses
- average amount by which scores differ from the mean
- more agreement
- the normal distribution
- ___ of all scores are ___ SD above or below the mean
- ____ of all scores are __ SD above or below the mean
- the bell shaped curve, mean and median identical
- 68%, 1
- 95%, 2
- deceptive statistics
- ___ vs ____ differences
- graphing data
- when not reported accurately, stats can be misleading
- relative vs absolute (double sounds like more than going from 1 to 2)
- need to check numbers on the axis
- Inferential statistics
- most common form =
- H0 vs H1
- making inferences about the population from sample
- null hypothesis significance testing *any procedure that produces a p value
- H0 = null hypothesis, no difference, effect being studied doesnt exist
H1 = alternative hypothesis, there is a difference
p value below alpha 0.5 =
p value above alpha 0.5 =
reject null hypothesis, statistically significant
fail to reject null hypothesis
*** P value
__ p values mean data is ____
Cannot conclude that H0 is ___ or ___ based on data
probability that these data could arise, if the null hypothesis were true
small, unlikely
likely or unlikely
p value replication
the p value you get the first time tells you nothing about what will happen the next time you measure p
- Trafimow said ….
P value myths: - 3.
4.
5.
- said NHST was logically invalid
- p < 0.5 = null hypothesis is unlikely
- rejecting null hypothesis and chance of false positive is 5%
- the results would replicate 95 out of 100 times
- there is a large difference or effect
- there is an important or significance effect
- Effect size
- significant
- nonsignificant
- Cohens d
- Measuring the strength of relationship between two variables
- p value lower than .05
- p > .05
- difference in means, in units of SD
Confidence intervals
our sample sizes can be estimates of population, but theres uncertainty about these estimates
CI are a quantification of this uncertainty
The larger the sample size
the narrower the CI and less uncertainty
- statistical power
- more power?
- ability to detect small effects using NHST
- Bigger samples
Bayesian statistics
provide estimates of whether null hypothesis is more or less likely based on data
SD of sample vs. population
sample = N-1
population = N
