Research Methods Final yuh

done with this shit

What are experimental within subjects designs also referred as

repeated measures design

What are the two possible experimental within subject conditions given to participants

either sequential: giving conditions one at a time, or same time:giving all the conditions at once

What is a threat to internal validity?

environmental variables: anything that can affect internal validity that occurs during the study

What’s a threat to between subjects designs but not within subject designs?

Individual differences are not a threats to within subjects designs

What is a history confound?

Something that occurs outside of the study from time 1 to time 2 that has to affect most of the people in the study

What is the maturation confound?

What is the instrumentation confound?

broken instrument used to measure data from time 1 to time 2

What is regression toward the mean?

The idea that if you were to get an extreme score at time one, your next score will be more likely to be what your usual score is or your more consistent score

What are order effects?

Anything where the first group affects the second group. This can include practice, fatigue, and carry over effects

How do we separate order effects from time related variables

Classification

How do we fix time related variables?

Not waiting as long to test second variable group

What would waiting a longer time between testing do?

Allowing for maturation effects and history effects to take place

What would wating a shorter amount of time cause?

Order effects such as practice or fatigue

How can all time related issues be solved?

Creating a between subjects design would eliminate all of the time-related confounding variables.

Wh

What are some problems with counter balancing?

just mixing in the variance, not fixing the confound, just mixing in results with each other. It makes it not a confound but it adds a lot more error variance.

Asymmetrical order effects

scores calculated differently based on the order in which treatments are administered. Counterbalancing cannot solve this issue

Whats a major issue with counterbalancing?

It is very hard to do a research study this way as if you have 4 levels, you now have to consider the possibility of 24 different conditions for all of the different possible combinations of order. You would also need a lot more people in the study as well as having to perform the individual analyses for all of the combinations.

What’s one positive about a within subjects design?

You can mathematically take out the individual differences by taking the individual averages

What does an F value consist of?

Within treatment variance as the numerator, between subjects variance (individual differences) on the bottom.

What are some advantages to a within subjects design?

Within subjects is more powerful and you can take out the individual differences

Also need less participants

What are some disadvantages to a within subjects design?

attrition, as you are asking the participants to do a lot more

What is a matched subjects design?

an experimental design where participants are matched on a relevant variable. cannot subtract out the individual differences but everyone is the “same”

What are non experimental designs and quasi experimental designs?

Randomly assigning and not controlling variables

has lower internal validity and you cannot determine cause and effect

Quasi experimental designs

appear to have a factor/independant variable. Usually looking for a relationship or a why something is occuring.

What are non equivalent groups?

didn’t randomly assign groups

Non experimental design strategy → Differential research designs

They have zero internal validity meaning you cannot determine why groups scores are different

Post test only non experimental design

Give group 1 the treatment and observe thier scores

Give group 2 nothing and observe thier scores

Cannot determine internal validity

Quasi experimental designs or pre test post test non equivelant group design

group 1, observe pretest scores, give treatment, then observe scores

group 2 observe pretest scores, observe non treatment scores

Higher internal validity but not as high as an experimental design

Within subjects design for non experimental and quasi-experimental

Cannot counter balance

Non experimental pretest post test

pretest and observe scores, provide treatment, then observe scores

an issue with thid design is that anything could have affected scores between pre and post tests

Quazi experimental design

Time series design menaing many observations over time

What is an interupted time series design

Something unexpectidly happened naturally

What some things that can occur when you don’t have manipulation or control?

it may be easier to do, but it may be unethical, and the logistics are trickier

What is the main problem with correlational research?

cannot determine causation

What are some other issues with correlational research

third variable problem, no manipulation or control

What are the three peices to a correlation

Direcitonality, shape and the strength or degree

How do you determine the strength for a linear relaitonship?

Pearson (r value) which goes from -1.00 to +1.00

What is spearman used for?

ordinal scale or where the difference between the scores are irrelevant and monotonic relationships

What is a monotnoic relationship?

A non straight line relationship where it is not increasing/decreasing at a constant rate and not consistent

What are some alternatives for when the data is nominal?

You can use the nominal data as a grouping variable by relabeling the groups with 0’s and 1’s. You can then tell the relationship between the nominal and non nominal variable.

What is the Chi-Square test?

grouping variables that have more than two levels. Chi square will determine if the data is statistically significant

What is the Phi-coefficient?

It only determines the strength of the relationship between two variables. Can only determine the strength between two groups at a time.

What is used for the determining the strength of relationships?

The coefficient of determination or r². You can convert r² into a percentage.

What does r² represent?

It represents the overlap between x and y or their shared variance. How much of the variable x is determined by y and vice versa.

What is statistical significance?

Determines whether the results were due to chance or not. r value is not statistical significance. Some issues with this could be that if you were to get a lot of people in your study, you could show that your results were not due to chance when they really were. This means you have to have a good reason for including a lot of people in a study.

What is a proxy?

A proxy is a variable used to substitute a variable you can’t get. Context matters as well as the variable might not be as strongly representative due to certain factors.

What is inter rater reliability?

All of the raters are saying the same thing

What is test retest reliability?

testing something more than once

What is split half reliability?

All of the questions in the first half of the test are related to the other half of the test

What kind of research (generally) are personality theories?

Mostly correlational research

What are some things that researchers due when finding a relationship between two variables?

providing an explanation for the results even though correlational research cannot be used to determine causation

What are the strengths of correlational research?

No manipulation or control needed. Logistically it is pretty easy as it just allows the researcher to measure scores. It also has very high external validity meaning you could simply measure scores from the real world.

What are some weaknesses with correlational research?

Cannot determine the direction (whether x effects y or y effects x). Third variable problem meaning there could be an extraneous variable which is actually causing a change in both x and y instead of x and y causing a change within each other. You cannot determine cause and effect. You also won’t have very high internal validity.

What is a partial correlation?

It is a relationship between x and y where you eliminate a factor that might affect the relationship.

What is multiple regression?

Multiple x’s predicting a y variable or outcome. If you already have certain variables, does another factor really matter for the outcome of y or are they just redundant and measure the same thing?