Task 8 Flashcards
What is error variance?
All the variance in the dependent variable caused by factors other than the independent variable
extraneous or subject related variables such as age, gender and personality
Example: Comparison of 2 depression treatments. Other influential factors could be: family support, age, diet, season, time of day, life events, personality traits…etc
How can you reduce error variance?
- by holding extraneous variables as constant as possible (e.g. experimental protocol)
- by matching subjects on influential characteristics (see Matched-Group or Within-Group design)
Name 3 other tragedies to reduce error variance
- choosing appropriate levels for the independent variable (Pilot-study)
- Random assignment of subjects
- Statistical analysis to determine the probability that error variance alone could cause differences between groups
Name 3 cases of confounding and their solutions
- biased sample =select random sample
- biased experimenter =blind study
- extraneous variable (which changes with independent variable) = careful planning of how confounding could play a role
Name 4 types of Between-Subject Designs
- Randomized Two-Group Design
- Randomized Multigroup Design
- Matched-Groups Design
- Matched-Pair Design
How does a randomized two-group-design looks like?
- you have a sample, random assignment
- -Group A gets Treatment 1
- Group B gets Treatment 2 - you get a mean for each treatment (mean1 & mean2)
Name 4 Pros and 2 Con for the randomized Two-Group Design
\+ requires few participants \+ no need for categorization \+ no need for pretesting \+ simple statistical analysis -no Information about type of relationship -limited sensitivity
What is a randomized Multigroup Design?
All most the same like a Two-group Design, with the difference that you have the opportunity to add groups, so you can add Treatments as well
Name 2 possibilities to add a group
- Quantitative change of the independent variable (Parametic Design) e.g. Cups of coffee 1,2, or 4
- Qualitative change of independent variable (Nonparametric Design)
e. g. Cups of coffee 1, 2 or tea
What is a Matched-Groups Design?
you have matched pairs which you split and randomly assign into Group A and Group B
Name 3 pros and 2 cons of Matched-groups Design
+matching characteristic is meaningful:
+ individual differences controlled for - higher sensitivity (less error variance)
+fewer subjects needed to show effect
- matching characteristic is not meaningful:
- power of statistical test is lower compared to completely randomized
Name 2 pros and 1 con for Matched-pair designs
What is it?
Matching into two groups
+ requires few participants
+ simple statistical analysis
- no information about type of relationship
What is a Matched Multigroup Design?
1 pro 1 con
is matched into several groups
can give information about the type of relationship
Problem: if there are more than 3 groups a within-subject design might be better
Explain what a within-subject design is
- Each subject undergoes all conditions
- No random assignment
- Also called “repeated-measures” design
What are the pros and cons of a within-subject design 3+ und 2-
\+ individual differences problem solved \+ very powerful \+ few participants - more demanding on subject - carryover effects - progressive error- general experience
name 6 sources of carryover
- learning: performance on task improves simply by repetition
- fatigue: performance on repeated task worsens, because of fatigue (Ermüdung)
- habituation: repeated exposure leads to reduced responsiveness (e.g. to stimulus)
- Sensitization: repeated exposure leads to heightened responsiveness
- Contrast: comparison of elements within the experiments can change their individual evaluation
- Adaption: a period of adaption can either increase or decrease responsiveness
there are 3 ways to deal with carryover effects, which are this?
- Counterbalancing
- Minimizing Carryover
- Statistical design
What are you doing by complete counterbalancing?
subjects are distributed over all different
sequencing options (eg ABC, ACB,BAC..etc)
N! if you have 4 conditions 4!
What are you doing by partial counterbalancing?
a random group of possible treatment orders is selected
what are you doing by Latin square?
each treatment is on each position equally often
as many options are chosen as there are treatments
Which 2 problems do you have by dealing with carryover effects?
Differential carryover effects: treatments have differently strong carry over effects
Irreversible changes: lasting improvement or damage
How can you minimize carryover?
name 2 ways
- Practice sessions: to eliminate carryover effects like Irreversible changes, Habituation or Sensitization
- Breaks: reducing short lasting Habituation effects, Adaptation or Fatigue
Which statistical design can you use to deal with carryover effects?
Factorial Design: treats treatment order as an independent variable
• Allows for accurate assessment and
consideration of carryover effects
What are the advantages of matching according to the lecture
- you can be certain the groups do not differ on the matched variable
- you can guarantee the presence of subgroups - external validity
What are the disadvantages of matching according to the lecture?
- groups are still heterogenous
- sometimes pretest required
- you can’t match on all variables
- with drop-out matching is not entirely correct anymore
What are you doing by matching?
divide variable with much influence equally over groups
What is blocking ?
=limited population
= keep suspect confounder constant
What is an advantage of blocking?
you have a homogenous group
What is a disadvantage of blocking?
you reduce your external validity
What is the Solomon Four-Group Design ?
is good to proof if their is a pretest sensitization or not
When to youse Between and when within?
Which question do you have to ask to find out?
- Is there an order effect?
if yes; is counterbalancing a solution?
-yes> check next consideration
-no > use Between - Does counterbalancing lead to more variance in the data than a between-subject design?
- no >use Within
- yes > use between - How many many participants can you test?
- How long does testing take?