Test
Sin título
Este test tiene tipos de pregunta que se abren en el reproductor de siempre.
Preguntas del test (21)
1Bootstrapping - Data it's used for: [blank_start]Small skewed[blank_end] data sets, statistics without a [blank_start]known distribution[blank_end] – can’t make inferences
Purpose: Can determine [blank_start]confidence intervals[blank_end] → assess whether dataset contains [blank_start]zero[blank_end] if doesn’t we can be confident there’s a [blank_start]difference between means[blank_end]- Adifference between means
- BSmall skewed
- Cknown distribution
- Dconfidence intervals
- Ezero
2Process of bootstrapping:
1. Randomly drawn subsamples are drawn from a larger sample over and over again ([blank_start]sample with replacement[blank_end]) (approx. 2000 times)
2. Scores are plotted on a [blank_start]frequency distribution[blank_end]
3. Determine if it looks [blank_start]normal or biased[blank_end]
4. Calculate [blank_start]confidence interval[blank_end] (2.5 & 97.5 as limits of 95%)
5. If it does NOT contain zero, can determine there’s a difference
E.g. paper sample (put paper in bag, take out one and put back, do 2000 times, put data in numerical order, take out percentiles)- Asample with replacement
- Bfrequency distribution
- Cnormal or biased
- Dconfidence interval
3Bootstrapping - Interpretation: If the confidence interval contains zero, there is [blank_start]no difference[blank_end] between the two groups
Problems: Can’t generalise results to problem to real world (only study), requires a reasonable sample size (>5), can’t correct for [blank_start]poorly designed[blank_end] studies and [blank_start]biased[blank_end] sampling from the population- Ano difference
- Bpoorly designed
- Cbiased
4Chi Sq 1 Way goodness of fit test: purpose - Tests how well an observed distribution of scores matches a distribution expected by [blank_start]chance[blank_end]
Process: Chance distribution = [blank_start]total counts/number of categories[blank_end]
Interpretation: The observed distribution of scores differs significantly from what would be expected by chance- Achance
- Btotal counts/number of categories
5Chi Sq 2 way contingency table
Purpose: Determines whether there is an [blank_start]association[blank_end] between 2 categorical/ordinal variables
If MEF is assumption is violated – use Kolmogorov-Smirnov test
- Effect size given by [blank_start]Kramer’s v[blank_end] (not 2x2) if 2x2 [blank_start]Fisher’s exact[blank_end]
Observed chi square is compared to critical value (df)
Independent = no effect
Interdependent = an effect- Aassociation
- BKramer’s v
- CFisher’s exact
6Kappa Test: Used to examine inter-rater reliability in situations involving 2 interdependent raters (estimates degree of correspondence between independent rates)
K = [agreement-chance]/1-chance
K = kappa statistic – want it to be [blank_start]higher[blank_end] –
[blank_start]7[blank_end] = great, [blank_start]5[blank_end] = good, [blank_start]3[blank_end] = fair- A7
- B5
- C3
- Dhigher
7Binomial Test - purpose
- AExamines whether two samples are drawn from populations with the same proportions of a given characteristic
(e.g. are males more likely to smoke than females) - BUsed to test for the probability of binary outcomes
(e.g. R vs. L turns) [use to estimate likelihood of getting numbers in random sample] - CTests the probability of event flow in equal sampling units
- Seeing where things occur most in an area (intervals of time/space i.e. number of customers coming in at time of day)
- AExamines whether two samples are drawn from populations with the same proportions of a given characteristic
8Proportion Difference Tests (parametric)
- ATests the probability of event flow in equal sampling units
- Seeing where things occur most in an area (intervals of time/space i.e. number of customers coming in at time of day) - BUsed to test for the probability of binary outcomes
(e.g. R vs. L turns) [use to estimate likelihood of getting numbers in random sample] - CExamines whether two samples are drawn from populations with the same proportions of a given characteristic
(e.g. are males more likely to smoke than females)
- ATests the probability of event flow in equal sampling units
9Poisson Test - purpose
- ATests the probability of event flow in equal sampling units
- Seeing where things occur most in an area (intervals of time/space i.e. number of customers coming in at time of day) - BExamines whether two samples are drawn from populations with the same proportions of a given characteristic
(e.g. are males more likely to smoke than females) - CUsed to test for the probability of binary outcomes
(e.g. R vs. L turns) [use to estimate likelihood of getting numbers in random sample]
- ATests the probability of event flow in equal sampling units
10McNemar test used for
- A3 time points
- B2 time points
11Cochran's Q - used for
- A2 time points
- B3 time points
12McNemar: [blank_start]Repeated[blank_end] test that examines whether the probability of a given event is more or less likely in the same sample over time ([blank_start]2 time points[blank_end]).
Cochran's Q: [blank_start]Repeated[blank_end] test that examines the probability of an event less likely in the [blank_start]same[blank_end] sample of people when assessed at a different point in time or in relation to a different choice ([blank_start]3 time points[blank_end] or more)
Can use [blank_start]together[blank_end]: Cochran’s Q for 3 time points to see if there’s any [blank_start]significance[blank_end], McNemar tests then [blank_start]compares[blank_end] the 2 time points to see where the significance lies
Tests indicate whether the likelihood of the binary outcome is differentially likely at different sampling points
Problems: - Can’t do a time x outcome chi sq. test because the data is measured at different time points and the data are non-independent (collected from the same people)- ARepeated
- B2 time points
- CRepeated
- Dsame
- E3 time points
- Ftogether
- Gsignificance
- Hcompares
13Logistic Regression: Dependent measure is binary
- Aim is to determine the best combination of variables which correctly classify cases according to their membership in the two categories defined by dependent measure (0,1) or (1,2)
1. Identify significant predictors
- [blank_start]Wald[blank_end] statistic (t-test)
- [blank_start]P value[blank_end]
2. Classification table identifies percentage of correctly classified cases (overall %)
3. Check wald ([blank_start]higher[blank_end]) and sig. ([blank_start]<.05[blank_end]),
Exp(B) = [blank_start]odds ratio[blank_end]
4. [blank_start]Odds ratio[blank_end] indicates the change in the p(E) resulting from a one unit change in the value of that predictor variable score – how much more likely are you going to do something with a unit increase- AWald
- BP value
- Chigher
- D<.05
- Eodds ratio
- FOdds ratio
14Simple Linear Regression
- ADetermines the lines of best fit
- BMinimises residuals by getting close to actual value
- CWant R squared to be smaller because indicates model explains variance
- DWant R squared to be larger because indicates model explains variance
- E70% is good
15Multiple Regression:
Stepwise: [blank_start]statistical logic[blank_end] (highest correlation first – forward, all in then take out ones without R squared dropping – backward, combines forwards and backwards - stepwise)
Hierarchical: [blank_start]theoretical logic[blank_end]
Standard: [blank_start]all variables at once[blank_end]- Astatistical logic
- Btheoretical logic
- Call variables at once
16Mediation
- AThe effect of A is carried to C via B
A → B → C
(C dependent, A and B predictors) - BAxB→ - the simple effect of A on C varies according to B
[Relationship between A and C differs depending on level of B]
- AThe effect of A is carried to C via B
17Moderation
- AThe effect of A is carried to C via B
A → B → C
(C dependent, A and B predictors) - BAxB→ - the simple effect of A on C varies according to B
[Relationship between A and C differs depending on level of B]
- AThe effect of A is carried to C via B
18Hierarchical Linear Modelling Mixed Models & Multi Level Modelling
Takes into account substructures, hidden subgroups, non-independent data, greater flexibility and versatility (captures subgrouping variance – can capture interesting contextual questions)
1. Many [blank_start]levels[blank_end] to MLM and cases can be [blank_start]nested[blank_end] within the levels – lower levels require more cases than higher levels
2. Level 1 = individual
Level 2 = class
Level 3 = [blank_start]year[blank_end]
Level 4 = school
Fixed effects: [blank_start]constant[blank_end] across individuals e.g. age, gender, religion – run by selecting all given and known levels of a variable (age etc.) and prediction based on variable (shown as intercept on graph for regression coefficients?)
Random effects: [blank_start]vary[blank_end] – test relationship between individual level predictors and dependent measure varies depending on the group they belong in – only sample some of the possible levels of a variable
Nature of relationships observed may differ depending on sub-groupings
- Overall results may lead to [blank_start]fallacy of composition[blank_end] (assuming that what is true of member is true for the whole group)
Intercepts and function slopes may differ- Alevels
- Bnested
- Cyear
- Dconstant
- Evary
- Ffallacy of composition
19Hierarchical Linear Modelling
Cross Sectional vs. Repeated Measures Analysis
CS – tests for [blank_start]hierarchical structures[blank_end] ([blank_start]HLM[blank_end])
RM – a type of [blank_start]longitudinal[blank_end] data analysis where repeated observations on individuals are considered data [blank_start]nested[blank_end] within the individuals (RM data level 1, individuals level 2)
[blank_start]CS[blank_end] – how much of the variance is explained by individual variation vs. differences between groupings?
[blank_start]RM[blank_end] – designed to avoid problems of covariance assumptions in traditional GLM RM
When model building, -2LLR is used to test for model improvements – test statistic gets smaller as model gets better
Overall fixed effects, specific fixed effects and reference categories
Covariance parameters (metric scores, age)
CS uses SPSS & RM uses long form and requires restricting. Each measurement has its own data line/people can appear multiple times- Ahierarchical structures
- BHLM
- Clongitudinal
- Dnested
- ECS
- FRM
20Variable types in MLM
- ASubject variable (ID)
- BFactors (gender)
- CCovariates (BMI, height, metric measurement)
- DRepeated factors
- EFixed vs. random
- FGender as a level
21POWER ANALYSIS
Tests for sample size (a [blank_start]priori[blank_end]), power levels ([blank_start]post hoc[blank_end]) and [blank_start]effect size[blank_end]
Can be used to determine the sample size required for research, evaluate study results inc. effect size and power a study had
Can determine any of the following as long as you have the other three:
- Power
- Alpha level (a)
- [blank_start]Sample[blank_end] size (N)
- Effect size (ES)
Useless outside of NHST framework, T1 & T2 error implications due to interpretation of NHST- Apriori
- Bpost hoc
- Ceffect size
- DSample
Información
- Descripción
- Test sobre Sin título, creado por Isabella Loxton el 19/06/2018.
- Autor
- Isabella Loxton
- Idioma
- English
- Visitas
- 0
- Valoración
- Aún sin valoraciones Valora este test