Karteikarten
DATA #2
Karten in diesem Stapel (75)
- 1What is a single case design?a research design where the subject serves as his own control rather than using another control group.
- 2what are the negatives about single case designs? what are the positives?negative: limited in terms of generalisability
the person acts as their own control
Not taking control measures at the same time
positives:
easy to get sample
can replicate findings on one person on another
can explore viability
maps well to clinical practice, easy to do in clinic - 3What are the 3 MAIN QUESTIONS we need to ask for ANY design1. Is the change real?
look at the P value to assess statistical significance
2. Is the change caused by the treatment?
look at internal validity, can you say the change was bought about by the IV? Also need to look at the type of design, AB, ABA, ABAB.
3. Is the change important?
need to look at effect size, need to look at clinical significant >.80 and look at external validity. - 4what is internal validity?to what extend can we conclude that the independent variable (treatment) has bought about changes observed in the dependent variable.
- 5Why do we need control in designs? How do we control for differences in a sample?- to establish internal validity. To show that it was the IV that bought about the change in the DV.
- we need to control for all other confounding variables before we can say that change is caused by the IV.
control in a sample through
randomization and having a large sample size. - 6In single case designs, the person acts as their own control group. What are 3 ways we control in this kind of setting?#1 TIME SERIES APPROACH
a baseline measure acts as the control before treatment begins
#2 USE UNTREATED ITEMS
can observe differences between treated stimuli and untreated stimuli. want to see no change in the untreated stimuli performance.
#3 HAVE AN UNTREATED MODALITY/BEHAVIOUR
use one mode or behaviour to target and treat and leave another as the control. - 7what are the strengths and weaknesses of a time series control?S: easy to administer baseline measure
W: perceptual judgements of a visual display showing baseline and treatment not very reliable. - 8what are the strengths and weaknesses of using a set of untreated items?S: forming a matches set is fairly easy.
Can see any differences between trained and untrained items.
W: the treatment effect is limited to the items used. - 9what are the strengths and weaknesses of a time series control? (don't know strengths)W: the severity of the deficit in both modalities needs to be the same which is hard to do.
Tasks need to be of similar difficulty - 10What are the two main approaches to single case designs?1. interrupted time series design
2. pre therapy/post therapy design - 11In time series designs why do you want a stable baseline?want a stable baseline which shows no change over time to show that you are measuring a consistent performance.
- 12What is an AB design? what are the limitations? (3)AB is the simplest design, a baseline measure is taken, then treatment begins.
LIMITATIONS:
ETHICAL PROBLEMS
holding off on treatment while you get a baseline
WEAK INTERNAL VALIDITY
don't know whether the treatment caused the effect, could have been time related factors like maturation or spontaneous recovery.
VISUAL DISPLAY OF CHANGE
is this enough to decide whether the change is real? high chance of error in perceptual judgement - 13What is a major problem with using visual analysis to determine if there has been change? What are 2 techniques that can help visual analysis?It is difficult to conclude that change and treatment are related just by looking. The reliability of making a perceptual judgement is poor.
we can add
1. Celeration lines which allow us to evaluate a trend. the slope of the line shows the trend.
2. Split middle line where we look if there is an equal number of values above and below the celeration line for the baseline phase. - 14What is the binomial test?If there is no change during treatment, we expect 50% of the scores to be above the line and 50% to be below.
We use a binomial test to see if the proportion above the line is greater or less than that occurring due to chance.
we want to binomial to be significant <.05 to tell us whether the obtained scores have a low chance probability - 15what test would you use in this case- comparing baseline and treatment scores?An independent t test to calculate differences between the means.
need to keep in mind that serial dependency increases the risk of making a Type I error. - 16Even with celleration lines, perceptual judgements can still be wrong. Statistical analysis is always better than visual analysis. What is the major problem with statistical analysis but?serial dependency- observations close together are correlated/similar compared to observations that are further apart. This increases type I error
- 17Describe the ABA design? what is an advantage?baseline, treatment, baseline
Has better internal validity because that second baseline measure can show whether changes were occurring only in that B treatment phase.
Can also evaluate treatment effect size by both baseline comparisons - 18Describe the ABAB design? what are the advantages?baseline, treatment, baseline, treatment
even better internal validity if changes are seen in between both baselines showing that changes occur with treatment and then can be reinstated the second time treatment occurs. - 19Describe multiple baseline design? Advantages?You give one person treatment while others stay on baseline. A difference between this person following treatment shows treatment had an effect. Then introduce second person, if they have similar results this increases internal validity as
1. results are replicated
2. only observed in treatment not in others - 20Discuss ABCB design? limitations?baseline, treatment 1, treatment 2, baseline
has poor internal validity as don't know which treatment is causing changes in baseline- is treatment 1 having more of an affect on results than treatment 2? - 21When we have single case design with nominal data... what design to we use? what are its components?pre- post treatment comparisons
O1 X 02
O1 = observation 1
X = treatment
O2 = observation 2 - 22what is a binomial VARIABLE?Is a variable with two mutually exclusive categories only.
- 23What is a binomial test? when do we use it?we use a binomial test when there is two possible outcomes.
By assuming a chance level, the binomial test tells is whether the client performed significantly above the chance level?? - 24If the dependent variable is binomial (2 exclusive categories) Identify appropriate analysis techniques for testing for significant differences between two independent proportions? (independent = sampling different participants) what are the assumptions/limitations of these tests?1. Z TEST OF TWO INDEPENDENT PROPORTIONS
assumptions/limitations: X cant be extreme, cant be too big or small
2. FISHER EXACT TEST
can be used when z test violated assumptions.
limitation: lacks power
3. CHI SQUARED TEST OF INDEPENDENCE - 25If the dependent variable is binomial (2 exclusive categories) Identify appropriate analysis techniques for testing for significant differences between two dependent proportions? (dependent = sampling same participants on 2 different occasions)1. McNemar statistical test
- 26what is the chi square goodness of fit?examines the likelihood that the difference between observed and expected frequencies has occurred by chance.
- 27what is the chi square test for independence?tests whether frequencies observed across 2 or more levels of one variable are independent of the level of the second variable
- 28what is the assumption for both Chi squared tests?Independence of cases
selection of one case should have no effect on the selection of others. - 29within single case designs that are pre/post test, what tests do we use to compare?wilcoxon signed ranks
mann whitney U - 30USING DIFFERENT PARTICIPANTS IS....INDEPENDENT/ BETWEEN GROUPS
- 31USING SAME PARTICIPANTS IS....REPEATED/WITHIN GROUPS
- 32What does the research hypothesis claim?that any changes seen are not due to chance and are real changes due to the effect of the IV on the DV
- 33What does the null hypothesis say?That no real difference exists between means and any differences observed are due to chance or sampling error alone.
- 34When do you use the independent T test (sample type/ data type)A parametric test
two independent samples and the data is interval or ratio - 35what are the assumptions of the independent t test?- interval or ratio data
- independent cases
-normally distributed population
- homogeneity of variance - 36If assumptions have been violated what test do you use?The mann whitney U
- 37When do you use the Mann whitney U test? (sample type/data type)two independent samples with ordinal data. or when assumptions of parametric independent t test have been violated.
Mann whitney uses the medians not the means. - 38what are the assumptions of the Mann whitney U?Only assumption is that the shape of the underlying distribution needs to be symmetrical about the median.
- 39When do you use the Paired Samples T test? (sample type/data type)to determine pre post test scores when the data is interval or ratio.
- 40What are the assumptions of the paired samples t test?- interval/ ratio data
-independently sampled cases
- normally distributed data
-homogeneity of variance - 41If the assumptions of the paired samples t test are violated what non parametric test do we use?The Wilcoxon matches pairs test.
- 42When do you use the Wilcoxon matched pairs test? (sample type/data type)when determining pre/post test scores when data is ordinal or when assumptions of parametric paired t test have been violated.
- 43what are the assumptions of wilcoxon?the distributions of different scores between conditions is roughly symmetrical
- 44What are cohens d for effect size when using t tests?.20 small
.50 medium
.80 large - 45What is a type 1 error?rejecting the null hypothesis when it is true
- 46what is type 2 error?failing to reject the null hypothesis when it is false
- 47what are 2 ways to increase the power of a study?ensure a large effect size
and
increase sample size - 48Why do we use 1.96 in calculating confidence intervals? *remember we don't want CI to overlap1.96 corresponds to a z score in a normal distribution that includes 95% of all cases about the mean of that distribution.
- 49When do we use a one way independent groups ANOVA?when there is a post treatment observation only (no pre test) and there are 3 or more independent groups (different people)
- 50what ratio does an ANOVA USE? what does the null hypothesis say about this ratio? what do we want it to be?ANOVA uses an F ratio. Null says F=1. So we want F >1 to reject the null
- 51how is the F ratio calculated?F = Between groups / within groups
- 52what are the assumptions of a one way independent groups ANOVA.- independent observations
-normally distributed population
-homogeneity of variance - 53what test do we use to look at normality ? what do we use to look at variance? do we want these tests to be significant?shapiro wilk
levenes
NO- want p > 0.05 - 54If the assumptions of a one way independent groups ANOVA are violated, what 3 options do we have?1. reduce alpha level
2. use welches test if variance is violated
3. use Kruskall wallis - 55What does a significant kruskal wallis tell us? what do we do when it is significant?significant tells us that at least two population medians differ.
Need to do a pairwise mann whitney test to find out exactly where the difference lies. Also need to do a bon ferroni correction to adjust the alpha level to controlf or TI error - 56Why is the F test an OMNIBUS TESTA significant F test is omnibus because it tells us that there is a difference between at least 2 means, but we don't know where this difference exists. We must to post hoc and pairwise comparison testing next.
- 57what does eta squared measure? what are the effect sizes?measures the proportion of variance in the DV that can be explained for by the IV. We want to see a large effect to say the IV is responsible for the variance.
.01 small, .06 medium, .15 large - 58what is omega squared a measure of?is an estimate of variance within the whole population that is explained for by the independent variable.
- 59What are post hoc tests? What are pairwise comparisons what do they compare? what are the issues? what are 2 examples? why are they not that powerful?post hoc tests are used after a significant F omnibus test.
used to tell us exactly where the difference lies by comparing pairs of means for a significant difference. Needs to maintain a low type 1 error rate across analysis so we do a bon ferroni correction.
eg) Tukeys
because we reduce the alpha level. - 60what are planned comparisons? what is an advantage of them?are used to test predicted differences. But the choice of what comparisons you test need to be scientifically driven. need to have a reason for wanting to compare.
you don't need a significant F to do so.
and by making limited comparisons you increase the statistical power. - 61what is an orthogonal comparison?comparisons are called "orthogonal" when the each comparison is among different groups. Comparing Groups A and B is orthogonal to comparing Groups C and D, because there is no information in the data from groups A and B that is relevant when comparing Groups C and D.
can use t tests to test. - 62what is trend analysis? how many trends can we test for?Allows us to predict what could happen in the future. We test for linear/qudradic or cubic relationships.
the number of trends we test for is 1 less than the number of conditions we have - 63What do we use ANCOVA?ANALYSIS OF COVARIANCE
with a significant F we want to conclude that any differences among means is due to the IV. however it is possible that groups actually differed before treatment began, and changes your seeing is due to this initial difference and not the IV.
This is called a confound. - 64how does ANCOVA work- what does it do?ANCOVA statistically controls the effects of unwanted confounding variables on the DV.
This is done by equalising the groups by adjusting the means so they no longer differ in terms of the covariate. - 65why might confounding variables exist in the first place?1. randomization was not successful
2. naturally occurring groups were used. - 66After ANCOVA if the groups still differ what does this mean? (if ancova is significant)that any changes you are seeing are because of the IV as the covariates have now been removed. we can reject the null hypothesis.
- 67what are the assumptions of ANCOVA? (4)1. A LINEAR RELATIONSHIP BETWEEN DV AND COVARIATE
2. NORMAL DISTRIBUTION
3. HOMOGENEITY OF VARIANCE
4. HOMOGENEITY OF SLOPES. - 68explain the difference between testing the null hypothesis for an IV in an ANOVA and ANCOVA.ANOVA examines the impact of the IV.
ANCOVA examines the impact of the IV whilst equalising conditions in terms of the covariate. - 69When do we use a REPEATED MEASURES ONE WAY ANOVA?When there is only ONE group of participants who undergo all levels of the independent variable where there are more than one level of that IV.
- 70what are the assumptions of a REPEATED MEASURES ONE WAY ANOVA? (3)- normality of the distribution
- homogeneity of variance
- sphericity - 71what does sphericity assume? what do we use when it is violated?that variance between all pairs is equal.
use hyun feldt. - 72what are the advantages of a repeated measure one way anova?- fewer subjects required
- increase power as individual differences not a problem as same participants used. - 73what are the dis advantages of a repeated measure one way anova?- carry over effects as same participants
- sphericity is hard to achieve, as measures taken close together are usually similar but measures taken further apart tend to differ and have increased variance. - 74if the assumptions of a repeated measure one way anova are violated, what test do we use?The friedman two way anova.
A significant Friedman means population
means are not the same. - 75If we have a significant Friedman two way ANOVA, what follow up tests do we use to find out exactly which medians differ?follow up with a wilcoxon test and reduce the alpha level through a bon ferroni correction.
Informationen
- Beschreibung
- Karteikarten am DATA #2, erstellt von LAUREN BELLINI am 11/11/2015.
- Erstellt von
- LAUREN BELLINI
- Aufrufe
- 1
- Bewertung