Test 2 Flashcards
What is the simplest experimental design?
Single factor design with two levels.
i.e. one Independent variable with two levels.
Is the simplest design but not often used by researchers because it more complex designs provided more informative conclusions and capture non linear relationships with between subjects design!!!
If IV of interest is MANIPULATED and participants are RANDOMLY ASSIGNED: what design is used?
Single factor Independent groups design
If IV is MANIPULATED and participants are MATCHED to avoid cofounding variables and then RANDOMLY ASSIGNED:
what design is used?
Single Factor Matched groups design
If IV is a SUBJECT factor, a BETWEEN SUBJECTS design, and equivalent groups are formed using MATCHING. What design is used?
Single Factor Ex Post Facto Design. If its between and subject variable it can only be an ex post facto design !
IF IV is MANIPULATED BY DEFINITION and uses a WITHIN SUBJECTS design: what single factor design is this?
And name an example study.
Single Factor Repeated Measures Design (e.g. Stroop effect studies)
Stroop effect study asked participants to read the names of colors when the ink colour was either the same or different to the colout being read. They found no significant difference in response time between groups. Their second study asked participants to name the colour instead of reading it. They now found a significant delay in response time when the name of the colour and colour of the ink did not match.
Used reverse Counterbalancing. A-B-B-A and B-A-A-B to control for order effects.
Data from Single Factor designs (1x IV with 2x Levels) are analysed with what inferential statistic?And What are its 3 assumptions?
T-Tests. If the following assumptions are meet:
(A) If data is interval or ratio
(B) Data is normally distributed
(C) There is homogeneity of variance
If IV has only 2 levels the results will always appear:
Linear, because the graph of results will only have two points.
Why use multilevel experimental designs? (2)
(A) To be able to identify non-linear relationships
(B) As a function of ruling out (or falsify) alternative explanations of the main result.
If Independent Variable is Discrete a ___ Graph should be used.
> Bar Graph, because there is not intermediate values exist.
Interpolation can not occur
Using a Line graph would be unethical as it distorts the findings presented and suggests there is a relationship between value points on the graph.
If independent variable is continuous a ___ graph can be used but a ___ graph is prefered.
A Bar Graph can be used but a Line Graph is preferred.
This is up to the researcher’s preference to decide which presentation of data will most effectively display the main findings of the study.
(3) Choices on presenting findings?
(A) Numerical results presented in sentences like a report. Is not ideal when there are more conditions or IV used because it becomes bothersome to read.
(B) Tables ideal for when you have specific values you wish to highlight to the reader (Means or SD’s) or if you have too many data points to make an effective graph.
(C) Graph is ideal if there is a non-liner relationship, relationship between variables or large difference between conditions.
Should you use both a table and a graph?
No. It is redundant to present the same findings in both table and graph form. The researcher needs to choose one format based on what will be most suited to their data and research question.
Multilevel Single factor designs typically are evaluated with ___ inferential statistics.
One Way ANOVA, analysis of variance. Assuming that:
(A) Interval or ratio data
(B) Normal distribution
(C) Homogeneity of Variance
3 assumptions made when using a t-test AND one way ANOVA analysis on single factor 2 level or multilevel designs?
(A) Interval or ratio data
(B) Normal Distribution
(C) Homogeneity of Variance
Using one way ANOVA. How do we identify if the results we found were significant?
F Ratio/Value. Generally, if its above 4 it is significant.
note: this only tells us that somewhere in the data there is a significant variance between groups (i.e.variances are different).
Using one way ANOVA. How do we identify where the significance difference lies?
Using a Post Hoc test to identify which means systematically vary as a result of IV and to a significant degree (0.05 %).
What are the two types of ANOVA analysis and what of the 4 designs do they each suit?
(A) One Way ANOVA
Best used for
Independent groups
designs and Ex Post
Facto designs.
(B) Repeated Measures
ANOVA
Best used for Matched
groups design and
Repeated Measures
designs.
What are the two types of t-tests and when should you use them?
(A) Independent Samples
T-Test
Best for Independent
groups design and Ex
Post Facto designs.
(B) Dependant Samples T-
Test
Best used for Matched
groups design and
Repeated Measures
design.(3) Types of Special Control conditions:
(A) Placebo
In latin “ I shall please”
Typically, when participants are led to believe they have been given or in the treatment condition when in fact they have been given a placebo (e.g. inactive pharmacological substance or subliminal dental tape).
Most common in drug research.
(B) Waitlist
When participants are in the control condition but told that they need to wait a couple of weeks before they can be given the intervention. Raises ethical concerns about making someone in need wait for treatment that could be beneficial.
Most common in assessing the effectiveness of programs or therapy.
(C) Yoked
When participants, for any reason, experience different events in the study.
Where procedural experiences of the control group participants are matched or correspond exactly to those of the treatment participants.
In order to keep the events constant, increase the researchers ability to compare the findings and rule out potential confounding variables.
Homogeneity of Variance:
One of the conditions that should be in effect in order to perform the PARAMATIC inferential tests (t-tests or ANOVA).
Refers to the facts that variability among all conditions should be similar (SD relatively equal)
Independent Groups design is a ___ Subjects design and its defining feature is ___ and is used to ___.
Between Subjects design and uses Random Assignment to create equivalent groups.
Matched groups design is a ____ subjects design, ___ and ___.
Between subjects design, matched and randomly assigned.
Difference between Matched groups and Ex Post Facto design?
Ex Post Facto designs can not use random assignment and their IV variable of focus is a subject variable.
Non-linear effects:
when the outcome does not form a straight line when graphed and can only be found in multilevel single factor designs where the IV has more than 2 levels.
If you Square the T value you get the:
F Ratio
What is the F Ratio
it examines the extent to which the obtained differences in mean could be due to chance or due to an other factor (like the IV-hopefully).
Is a ratio of variances: variation between sample means/varitation within samples
We use an f distribution to calculate probabilities. Using the numerator and denominator of kir f ratio we compare it to our critical f table to plot onto the f distribution. If it falls within the rejection region we can conclude that the sample means are not equal and reject the null hypothesis
Are larger F ratio means there is more variance I.e. a measure of dispersion how far the data is scattered from the mean.
Mean squares are an estimate of population variance that accounts for df used to calculate that estimate.
Why can’t you conduct multiple t-tests to analyse multilevel single factor designs?
Because as you increase the number of times you repeat a statistical analysis you also increase the chances of you making a type one error- false positive, saying there is a difference between means when there isn’t.
The solution is using ANOVA test as your inferential analysis.
Types of Post Hoc Tests
(A) Non correlational
Very liberal post hoc test which is equivalent to conducting multiple t-tests without correction. Meaning the chances of making a type 1 error are very high.
(B) Tukey ☆
Does a minor correction to the t-distribution based on the number of comparisons being made in an attempt to keep the alpha in check.
(C) Scheffe
Very conservative test where there is never a good reason for using a scheffe post hoc test because the chances of making a type 2 error are very high- failing to report a significant difference in sample means.
(D) Bonferonni ☆
Similar to the tukey post hoc test but the Bonferonni corrects for type 1 error by dividing the significance level .05 by the number of comparisons being made. This is so when all the tests have been conducted the overall sognificance level will be .05 and keeping the chance of the results being due to chance at less than 5%
Defence for the use of control groups:
(A) We can not identify what treatments are effective till after a sufficiently designed experiment is conducted.
(B) People are not denied treatment, they are placed on a waiting-list and given it post study and during the study people are not denied the current treatment they have.
(C) they cost money and identifying the most effective treatments are beneficial for peoples well being and economically.
- These special control groups are more informative when using multilevel single factor designs.
Counterbalancing for Within subjects multi level designs:
Counterbalancing becomes much more complex:
(A) Latin square
prone to carry over
effects and random
error confound.
Implicated by the
resources required
and sample size used.
e.g. A B C D E
B C D E A
C D E A B
D E A B C
E A B C D
(B) solve this with the “Balanced Latin Square”.
Counter balancing for Single factor design with 2 levels:
Limited to AB and BA.
3 types of counterbalancing for within subjects single factor designs I.e. 1 IV with 2 levels
- Complete Counterbalancing
When participants only experience each condition ONCE.
I.e. 1/2 participants experience A-B and then other 1/2 experience B-A.
- Reverse Counterbalacing
When participants experience each condition MORE THAN ONCE.
I.e. A-B-B-A and B-A-A-B
- Alternate Counterbalancig
When participants experience each condition MORE THAN ONCE.
I.e. A-B-A-B and B-A-B-A
Random assignment is used for….
create equivalent groups.
Dependant samples T Test is also called? And what single factor designs can they analysis?
Paired Samples T-Test.
Used when the design is Within subjects (I.e repeated measures) or when there is a relationship between the two independent groups (i.e. Matched Group).
What non parametric test can be used to analysis the difference between two group means IF the assumption of homogeneity of variance has been violated?
Marnn-Whitney’s U test which compares two independent group means, much like an independent samples t- test, but without assuming they’re both normally distributed.
Magnitude of difference can be assessed by:
Effect Size.
Concern for Ex Post Facto designs?
External Validity. Matching is used to allow researchers to be more confident that the difference in means is due to the IV subject factor and not other confounds/existing differences.
Ex Post Facto Designs used to investigate what topics or research themes?
》Ex Post Facto designs are typically used to study natural disasters impact on human behavior.
》It’s unethical to manipulate this as an IV so researchers take advantage of naturally occurring disasters and an ex post facto design.
》Example: Sibbley & Bulbia (2012)
Who looked at the effects of the Christchurch earthquake on people’s tendency to rely in religon to cope with loss and pain.
▪︎IV (subject variable matched) whether they lived in Christchurch or anywhere else in NZ.
▪︎outcomes were coded into 4 groups: believers, disbelievers (who maintained their original beliefs), converts (were not before but were after) and apostate (were before and were not after).
▪︎ found no difference in believers or disbelievers
▪︎ an increase in the number of converts and apostates in both conditions but higher with those in Christchurch
▪︎concluded that subjects in Christchurch were more likely to turn to region to cope with emotional distress post earthquake and conversely more likely to leave religon as well compared to anywhere else in NZ.
