Ch. 9 - Research & Program Evaluation Flashcards
Internal validity
- refers to whether the DVs were truly influenced by the IVs or whether other factors had an impact
External validity
- refers to whether the experimental research results can be generalized to larger populations
Experiments emphsaize parsimony, which means
- interpreting the results in the simplest ways
Occam’s Razor suggests
- that experimenters interpret the results in the simplest manner
Flaws in research are often called
- bubbles (think air bubbles stuck under a sticker)
An experiment is confounded when…
- undesirable variables are not kept out of the experiment
- AKA contaminating variable
Basic research vs. applied research
Basic research: conducted to advance our understanding of theory
Applied research AKA action research: conducted to advance our knowledge of how theories, skills, and techniques can be used in terms of practical application
Causal-comparative design
- a true experiment that lacks randomly assigned groups
- data can be analyzed with a test of significance (i.e., t-test or ANOVA) just like experiement
If you can’t randomly assign subjects into groups…
- considered quasi-experimental
Hypothesis testing is related to work of
RA Fisher
Null hypothesis
- suggests that there will not be a significant difference between groups
- the IV does not affect the DV
Alternative hypothesis
- suggests that there will be a significant difference between groups
- IV does affect DV
Percentage vs. percentile
Percentage: Raw score
Percentile: descriptive statistic that thells what percentage of cases fell below a certain level
p = .05 means…
- differences really do exist
- will obtain samee results 95/100 times
- 5% error factor
Type I error
- When you reject the null when it is true
- lowering significance levels lowers type I errors
- AKA alpha error
Type II error
- When you accept the null when it is false
- AKA beta error
- lowering significance raises the risk of type II error
If researcher changes significance level from .05 to .001
- type I error decrease, type II error increase
t-test
- used to compare two groups with single IV
one-way ANOVA
- used when there is more than one level of a single IV
two-way ANOVA
- used with 2 IVs
- more than 1 IV = factorial design
Correlation coefficient
- indicates degree or magnitude of relationship between two variables
- degree of linear relationship
N = 1
- intensive experimental design
- case study
- popular with behaviorists who seek overt behavioral changes
- AKA idiographic studies or single-subject designs
Normal distribution stats
- 68% fall within +- 1 SD
- 95% fall within +- 2 SD
- 99.7% fall within +- 3 SD
Regardless of the shape the _ will always be the high point when a distribution is displayed graphically
- mode
X axis used to
Y axis used to
- plot the IV (horizontal)
- plot the DV (vertical)
Range
- highest - lowerst score or highest - lowest score + 1
- increases with sample size
Z-scores
- same as SDs
- AKA standard scores
Platykurtic distribution
- looks like the upper half of a hot dog lying on its side
- flat and spread out
Kurtosis =
- refers to the peakedness of a frequency distribution
Leptokurtic =
- tall and thin distribution
Stanine scores
- divide the distribution into nine equal intervals with stanine 1 as the lowerest ninth and 9 as the highest ninth
Nominal
Ordinal
Interval
Ratio
- Qualitative; distinguish groups; no true 0; does not indicate order
- Rank order; relative distance not always equal; relative placement or standing; does not delineate absolute differences
- Interval: numbers scaled at equal distances; no absolute 0; can add and subtract, but can’t multiply or divide; IQ tests are interval
- Ratio: Interval scale with true 0; addition, subtraction, multiplication, division possible; most psychological attributes cannot be measured on ratio scale
Nocebo
Placebo
- has a negative effect
- i.e., when a doctor comments one only has 6 weeks to live
- placebo has possitive effect
Hawthorne effect
- if subjects know they are a part of an experiment, their performance sometimes improves
- reacting to the presence of the investigation
- AKA observer effect
Rosenthal effect
- AKA experimenter expectancy effect
- the experimenter’s beliefs about the individual may cause the individual to be treated in a special way so that the individual begins to fulfill the experimenter’s expectations
Halo effect
- when a trait which is not being evaluated influences a researcher’s rating on another trait (i.e., attractiveness and counseling)
Statistical regression
- predicts that very high and very low scores will move toward the mean if a test is administered again
Ipsative implies
- within-person analysis (i.e., was you jog faster today than yesterday) rather than a normative analysis between individuals
Demand characteristics
- can confound an experiment
- relates to any bit of knowledge - correct or incorrect - that the subject in an experiment is aware of that can influence their behavior
Post-hoc tests
- Duncan’s multiple-range
- Tuey’s
- Scheffe
- further discriminates between the ANOVA groups
Pygmalion effect
- Rosenthal effect and experiment becomes a self-fulfilling prophecy
Ahistoric therapy
- any therapy that focuses on the here and now
Statified sampling
- When a special characteristic needs to be represented in the sample, such as race, gender, etc.
Snowball sampling
- when subjects invite others to the study
Sampling error
- small samples do not mimic the population
Systematic sampling
- picking a number between 1 and 10 and then picking every nth person
Operational definition
- outlines a procedure
- operationally define procedures so one can replicate the procedure
Axiom
- universally accepted idea needing no additional proof
Non-parametric measures
- Mann-Whitney U test
- Wilcoxon signed-rank test for matched pairs
- Soloman and the Kruskal-Wallis H test
Matched design
- subjects are matched in regard to any variable that could be correlated with the DV, which is really the post-experimental performance
- unmatched groups are known as independent groups
Inductive logic or reasoning
Deductive logic or reasoning
- the research goes from specific to generalization
- research goes from general to specific
Standard error of measurement (SEM)
- tells what would most likely occur if the same individual took the same test again