2017-11-07 02 Exam Flashcards
Normal distribution and skewness
(Manual, 35)
- A skewness between -2 and +2 is normal
scales of measurement
(in-class 9/4) (A &; L, 79-83)
- nominal
- ratio
- interval
- ordinal (not needed)
nominal scale
(in-class 9/4) (A & L, 80) -used to measure categories
ratio scale
(in-class 9/4) (A & L, 80)
- true zero (fixed-point)
- used to measure quantities
interval scale
(in-class 9/4) (A & L, 80)
- used to measure ratings
- identity (each number has a specific meaning), order (numbers on a scale, in ordered sequence), equal intervals (distance between numbers on the scale is equal)
operational definition
(in-class 9/4) (A & L, 77)
- specifics of how the variable is measured
- so it can be exactly replicated
central tendency
(A & L, 147)
- central score
- summarizes center of distribution
- mode, median, mean
mode
(A & L, 149)
measures central tendency
- most frequent score in a distribution
median
(A & L, 149)
measures central tendency
- halfway point of distribution
mean
(A & L, 149)
measures central tendency
- arithmetic average
variability
(A & L, 150)
- how much scores are different from each other in a sample
- observed minimum, observed maximum, range, standard deviation
observed minimum
(A & L, 150)
measures variability
- lowest score in the sample
observed maximum
(A & L, 150)
measures variability
- highest score in the sample
range
(A & L, 150)
measures variability
- distance between observed minimum and maximum
standard deviation
(A & L, 150)
measures variability
- how much in general the scores in a sample differ from the mean
descriptive statistics
(A & L, 142)
- used to analyze quantitative and qualitative data
- quantitative analysis used to summarize characteristics of a sample
- CT: mode, median, mean
- Variability observed minimum, observed maximum, range, standard deviation
descriptive statistics for nominal data
(A & L, 170)
- frequencies and/or percentages
- CT: (sometimes mode)
- variability: –
descriptive statistics for interval or ratio (normal distribution)
(A & L, 170)
- (sometimes: percentages for each score on an interval scale)
- CT: mean
- variability: standard deviation (sometimes: possible min/max for interval, observed min/max for interval and ratio)
descriptive statistics for interval or ratio (skewed)
(A &; L, 170)
- (sometimes: cumulative percentage)
- CT: median
- variability: observed min/max or range
sampling
(A & L, 119)
- process of how the sample is selected
probability sampling (random sampling)
(A & L, 121)
- sampling procedure that uses random selection
- ideal, (external validity/generalizable)
- simple random, stratified random, cluster sampling
non-probability sampling (non-random sampling)
(A & L, 123)
- sampling procedure that doesn’t use random selection
- less time (no need to identify all participants [members, clusters] in a population)
- if researcher can’t identify all members/clusters, appropriate sample size, and/or minimize non-response data
- convenience, quota, maximum stratification, snowball,
convenience sampling
(A & L, 129)
non-probability sampling
- sample is volunteers who are readily available and willing to participate
- typically have an over-represented group
- easiest (feasable)
snowball sampling
(A & L, 132)
non-probability sampling
- participants recruit others into the sample
independent variable (and levels)
(A & L, 21) (in-class 8/29)
- variable that’s manipulated in an experiment
- Levels: a control group and then 1 or more other assignments/groups
dependent variables
(A & L, ) (in-class 8/29)
- variable that’s measured in an experiment
- expected to change based on IV
pilot study definition
(in-prac 9/18)
- still with target population
- test before spending money
- work on any possible changes
Pilot studies can find problems with
(in-class 10/24)
- recruitment
- retention (who will stay?)
- implementation (measures good?)
- assessment (is it accurate?)
- new methods (money)
Experimental design
(A & L, 19) 1 - random assignment 2 - IV manipulated (at least 2 levels) 3 - DV measured Main benefit: can determine causality
Random assignment definition
(A & L, 280, 184)
- essential for an experiment
- participants (already selected) chosen at random to IV conditions/levels
Random assignment and purpose
(A & L, 280, 284)
- increases internal validity
- IV groups to be as similar as possible before IV exposure
- evens out individual differences across IV conditions
(in-class 10/26)
- any group differences between groups isn’t the confounds (confounds affect both groups)
Independent variable (and levels)
(A & L, 21) (in-class 8/29)
- variable that’s manipulated in an experiment
- Levels: a control group and then 1 or more other assignments/groups
IV manipulation: reliable and valid
(A & L, 308)
- need equivalent IV levels/conditions
- manipulation check
Manipulation checks
(A & L, 292)
- Checking if what you manipulated what you wanted to manipulate
EX: if part of the study was to read a book, quiz participants on their comprehension of the book
Pilot studies can find problems with
(in-class 10/24)
- recruitment
- retention (who will stay?)
- implementation (measures good?)
- assessment (is it accurate?)
- new methods (money)
Confounding variables definition
(In-class 10/24) - History effect - Maturation effect - Testing effect - Instrumentation effect - Regression to the mean (statistical regression) (( usually more than one at once ))
Confounding variables: ways to limit
(In-class 10/24)
- random assignment
- manipulate ONE variable
- need equivalent IV levels/conditions
- large sample size
History effect
(In-class 10/24)
- (due to experiences or environmental factors)
- changes may be due to outside events
- anything external to the study
- if only affecting one IV group (as average), then probably history confound
Maturation effect
(In-class 10/24)
-(due to experiences or environmental factors)
- changes due to participants’ internal changes over time
- more likely across long time periods or with young children
(kind of opposite history confound)
Testing effect
(In-class 10/24)
-(due to experiences or environmental factors)
- repeated testing can impact results
EX: students used to professor’s tests, not knowledge growth
Instrumentation effect
(In-class 10/24)
-(due to experiences or environmental factors)
- changes in measurement instrument can cause changes in DV
EX: measuring children < 3y/o on a table, and > 3 y/o standing
Regression to the mean (statistical regression)
(In-class 10/24)
- (due to participant characteristics)
- scores that are selected because they’re extreme are likely to be less extreme when retested
Threats to internal validity
(In-class 10/24)
- Confounds:
- History effect
- Maturation effect
- Testing effect
- Instrumentation effect
- Regression to the mean (statistical regression)
Criteria for causality
(A & L, 271)
- Correlation: (relationship between A and B)
- Sequence: (change in A comes before change in B)
- Ruling out confounds: (controlled for possible confounds, so A must be the only factor to cause change in B)
Inferential statistics definition
(A & L, 185)
- statistical analysis of data from one sample to draw conclusions about population sample is from
Descriptive statistics vs inferential statistics
Descriptive statistics: depend on skewness and scale of measurement (Central tendency and variability)
Inferential statistics: depend on scale of measurement and levels of IV
Null hypothesis
(A & L, 190) (in-class 10/26)
- prediction of no difference between groups
- don’t assume reader already knows IV, DV, or levels
EX: “no difference” “similarly”
Alternative hypothesis (directional)
(A & L, 197)
-(one-tailed)
- prediction of the direction the results from a sample will differ from the population
EX: “better than” “highest”
Alternative hypothesis (non-directional)
(A & L, 197)
-(two-tailed)
- prediction that results from a sample will differ from the population without saying how
EX: “there will be a difference”
Alternative directional hypothesis for multi-level
(in prac 10/30)
- mention DV
- mention IV levels and how they’re related to each other
EX: “A will be more than B” “followed by”
Type I error
- When rejecting NULL
- less than (
Type II error
- When retaining NULL
- more than (>) .05
- can never know chance
Reject null
- Type I error
- chance is p value ( less than (
Retain null
Type II error
when p is more than (>) .05
- can never know chance
Power definition and impacts
(A & L, 205)
- Ability to correctly reject the null hypothesis
- Factors:
- sample size
- amount of error
- strength of effect
Power: how to increase
- increase sample size
- increase effect size
- increase within-group homogeneity
- increase between-group heterogeneity
Between groups variance (treatment variance)
(A & L, 330)
- variability between groups/levels/conditions
- want to MINimize this
Within groups variance (error variance)
(A & L, 330)
- variability among participants scores (in same group/level/condition)
- want to MAXimize this
Pearson’s r definition
- correlation coefficient that tells magnitude of relationship between 2 variables (r^2)
- (interval/ratio) and (interval/ratio)
Pearson’s r assumptions
interval/ratio and interval/ratio
Pearson’s r analysis
- scatter plot
- SPSS: “correlation coefficient”, p value: “Sig. (2-tailed)”
Pearson’s r effect size
Pearson’s r is the effect size (tells magnitude of relationship) between 2 variables (r^2)
- small: r ~ .1, r^2 ~ .01 (1% variance accounted for) (( absolute value ))
- medium: r ~ .3, r^2 ~ .09 (9% variance accounted for) (( absolute value ))
- large: r ~ .5, r^2 ~ .25 (25% variance accounted for) (( absolute value ))
Pearson’s r formula for results section
-SPSS: “correlation coefficient”
- p value: “Sig. (2-tailed)
“ (r = ._ _, p = . _ _ _) “
Chi-square test of independence definition
- (nominal) and (nominal)
- examines distribution frequencies
Chi-square test of independence assumptions
- (nominal) and (nominal)
- independent groups ( no matching/repeated measures)
- expected frequency of at least 5 in each cell
- variables not related to each other
Chi-square test of independence analysis
- SPSS effect size: “Phi” in “Symmetric Measures” ( phi-squared ( ϕ^2 ) )
Chi-square test of independence effect size
- phi-squared ( ϕ^2 )
- small: ϕ^2 ~ .1, r^2 ~ .01 (1% variance accounted for)
- medium: ϕ^2 ~ .3, r^2 ~ .09 (9% variance accounted for)
- large: ϕ^2 ~ .5, r^2 ~ .25 (25% variance accounted for)
Chi-square test of independence formula for results section
- “ X^2(df, N = #) = “Value” under “Pearson Chi-Sq.”, p = ._ _ , ϕ^2 = . _. ”
Independent-samples t test definition
- (nominal grouping/dichotomous) and (interval/ratio)
Independent-samples t test assumptions
- groups are independent
- IV (or grouping) nominal grouping/dichotomous
- DV (or outcome) is interval/ratio
Independent-samples t test effect size
Cohen’s d (formula given)
- small: d ~ .20
- medium: d ~ .50
- large: d ~ .80
Squared point biserial correlation rpb^2
- gives percentage of variance of outcome (DV) accounted for by predictor (IV)
Independent-samples t test formula for results section
“ t(df) = t#, p = ._ _ _, d = . _ ”
P-value
- less than () .05 : not stat. sig., retain null, chance of Type II error
Statistically significant
- when p is less than (>) .05, reject null, chance of Type I error
NOT statistically significant
- when p is more than (>) .05 : not stat. sig., retain null, chance of Type II error
Homogeneity of variance
(A & L, 316)
- assumption that variance of populations is the same
- group SDs are estimates of the population variances
Levene’s test
In independent-samples t test
- Not Stat. Sig. (p ≥ .05), SECOND line
- Stat. Sig.(p ≤ .05), FIRST line
Effect size definition
describes strength/magnitude of IV effect
for simple experiment with independent groups
Effect size: r
Pearson’s r is the effect size (tells magnitude of relationship) between 2 variables (r^2)
- small: r ~ .1, r^2 ~ .01 (1% variance accounted for) (( absolute value ))
- medium: r ~ .3, r^2 ~ .09 (9% variance accounted for) (( absolute value ))
- large: r ~ .5, r^2 ~ .25 (25% variance accounted for) (( absolute value ))
Effect size: phi-squared ( ϕ^2 )
Chi-square test of independence
- small: ϕ^2 ~ .1, r^2 ~ .01 (1% variance accounted for)
- medium: ϕ^2 ~ .3, r^2 ~ .09 (9% variance accounted for)
- large: ϕ^2 ~ .5, r^2 ~ .25 (25% variance accounted for)
Effect size: Cohen’s d
Independent-samples t test
(formula given)
- small: d ~ .20
- medium: d ~ .50
- large: d ~ .80
Squared point biserial correlation rpb^2
- gives percentage of variance of outcome (DV) accounted for by predictor (IV)
Quasi-experiment definition
1 - NO random assignment
2 - IV manipulated (at least 2 levels)
3 - DV measured
Quasi-experiment advantages
When unethical to have random assignment (participant age, gender)
Multi group (multi level) experiment definition
- IV with 3 or more levels
- DV as usual (measured)
Multi group (multi level) experiment advantages compared to many simple experiments
- decreases probability of a Type I error
- decreases confounding
- increases efficiency (decreases # of studies and participants)
- increases internal validity (examines functional relationships, and non-linear/linear)
Multi group (multi level) experiment limitations weigh advantages and disadvantages with topic
- what are the research questions/hypothesis?
- what would be the advantages for this study of adding a 3rd condition?
- what would be the disadvantages?
Results section APA
- heading centered, double spaced, indented paragraph
- M, SD of participants
- inferential statistic test
- formula for inferential statistic
- statistically significant?
Discussion section APA
- restate hypothesis
- statement about meaning/implications of results
- limitations
- new directions
Multi group (multi level) experiment assumptions
- IV with 3 or more levels
- DV as usual (measured)
Simple experiment definition
- IV: manipulated, 2 conditions, nominal
- DV: interval/ratio
Simple experiment advantages
- simple (relative to multiple group)
- maybe smaller sample size
Simple experiment limitations
- only 2 groups
- non-linear
Practical significance
(A & L, 209)
- usefulness and everyday impact