Research methods C Flashcards
Types of rationale (4)
From previous researches methodological problems
Considering different theories to explain past research
Replication
Unique theory based on general observations
Types of design (3)
Simple comparisons - one IV, two conditions
One way designs - one IV, three or more conditions
Factorial designs - more than one IV, multiple simultaneous experiments
Example of factorial design
Effect of rain (IV 1) and wind (IV 2) on perceived pleasantness (DV)
How many interactions and conditions are there with two factors
One interactions and 4 conditions
How many interactions are there with 3 factors
4 interactions, one being all of them together
What does an ANOVA do
Is an analysis of the variance (SD^2), it determines if two or more groups are from the same population of scores
What do ANOVAs compare… and if they are similar?
The within group error variance to the between group error variance
… if they are similar then the groups are from the same population
When might ANOVAs be significant
If the between group variance is substantially larger than the within group variance
Why do within group designs tend to be more sensitive to ANOVAs than between group designs
Has smaller error variance so more likely for the between group variance (caused by IV) to be significantly larger than within group variance
(assuming minimal carry-over effects)
Sources of variance (2)
Error - different people producing different scores
Effect of variables / different conditions
F-ratio =…
Between group variance / within group variance
General rule for F-ratios showing significance
If they Re greater than 1, most likely significant
How to read F-ratios
The larger the more significantly the between group variance is larger than the within group variance
What a significant ANOVA tells us and what it doesnt
At least one group is significantly different from at least one other
Doesn’t tell us which groups
How to determine which groups significantly differ after a significant ANOVA
Comparing means And SDs indicates which ones
But post-hoc comparisons to find simple effects does so statistically
Problem with post-hoc comparisons
Familywise error rates
How to carry out post hoc comparisons
Use the appropriate t-test to compare each separate pairs of conditions
Type 1 error
Concluded significant when it is not
Went to find wolf but there wasn’t one
Type II error
Conclude no significance when there is
Didn’t go to wolf but there was
Explain family wise errors
Chance of type I errors = p, if comparing 3 groups then 3 comparisons needed… what would usually be 1/20 (.05) is now 3/20 (.15), meaning significance threshold too high
P =.25 threshold if 5 comparisons (**though threshold would still say .05!!)
How to avoid family wise errors
Use bonferroni correction - divide .05 by number of comparisons to make new criterion for significance
parametric assumptions (3)
normally distributed
Homogeneity of variance
Interval or ratio data
If one is not met, a non parametric test is needed
Limitations of non parametric tests (2)
Provide limited information (no means or variance)
Greater chance of type II errors as less sensitive
How to report non parametric tests (3)
State that (and how) assumptions for parametric tests were violated Report medians and range information (maybe IQR) Report test statistics