aaaaaa

Levels of a Factor

Individual conditions/values that make up a factor

Ex:

1. therapy technique (2 techniques)

2. time (three diff times)

F-ratio

variability between samples means

-----------------------------------------

variability expected by chance (due to error)

- Based on variance

- no systematic treatment effects = 1.00

- systematic treatment effects = substantially larger than 1.00

- great indicator of existence of said effects

Error Term

This provides a measure of the variance caused by random, unsystematic differences

- denominator of f-ratio

- when treatment effect is zero and Ho is true, f-ratio ends up close to 1.00

Turkey's HSD

Calculating how big of a diff the between means has to be that is statistically significant for study

- compares each observed mean difference with this and determine which pairs of means differ

Between-Groups Variance

Ultimately measuring the differences between sample means

ex: m1=5, m2=13, m3=18

- one-way: treatment effects, ind. diff, exp. error

- repeated measures: treatment effects, exp. error

Within-Group Variance

Ultimately measuring the variability inside each condition

x3= {16, 18, 20}

- one-way and repeated measures: ind. diff and exp. error

- as this increases, the f-ratio will decrease

Factorial Designs

Experimental designs with more than one IV (or factor)

Ex: Being invited to a party -- location, people invited, time

Main Effects

Effect of a single factor averaging across all other factors

Interactions

Effect of one factor differs across levels of another factor

Effect sizes/Eta Squared

Measures the percentage of variance accounted for by treatment (manipulation)

- .01 < n2 < .09 small effect

- .09 < n2 < .25 medium effect

- n2 > .25 large effect

Simple Main Effects

Effect of one independent variable at any given level of another

ANOVA

Purpose is to evaluate differences in means between treatments

F-Ratio for One-Way Analysis

MSbetween

--------------

MSwithin

F-Ratio for Repeated Measures

MSbetweengroups

----------------------

MSerror

Post-Hoc tests

to determine which means are significant and which aren't

- used after ANOVA when you reject Ho and there's more than three conditions (k>3)

Experimenter bias

Extent to which the behavior of a researcher or experimenter intentionally or unintentionally influences the results of the study

To reduce experimenter bias

• Get a second opinion

• Standardize the research procedures

• Conduct a double-blind study

get second opinion

have other researchers involved

standardize research procedures

the more we have things be the same across people, the better it will be

--> interrater behavior

--> same ratings/scoring of measures

--> more open to interpretation creates more ability for bias

Double-blind study

A type of research study in which the researcher collecting the data and the participants in the study are unaware of the conditions that participants are assigned

--> label drugs with anonymous markers (ie. 1/2 or A/B)

Sensitivity of a measure

Extent to which a measure can change or be different in the presence of a manipulation

--> assume there will be a difference/change, but that may vary

--> if will produce a range in scores

--> if they don't vary much, something is wrong

example of sensitivity

blood pressure/heartrate fluctuate based on reaction

--> assume different videos will impact heartrate

--> if you show video and there is no change, there is no sensitivity

--> have to figure out how to produce a change

Range effects

Limitation in the range of data measured in which scores are clustered to one extreme

--> possible range of scores show minimal change that is pegged to one end or the other (can have all really low scores or all really high scores)

example of floor effect (range effects)

if you are measuring memory . . .

if everyone gets all the questions wrong = shows no various because the test is too hard

--> no range/ability to see good/poor because everyone is poor

floor effect

when scores are clustered very low; measure is too difficult

Ceiling effect (range effect)

scores are clustered very high; measure is too easy

example of ceiling effect

if measuring memory. . .

if all participants get everything right, there is no different between the two groups

We need scores

that are different across groups/people

--> the less we see variability, the less we know because it doesn't tell us anything

To maximize the sensitivity of a measure and minimize range effects

• Perform a thorough literature review

• Conduct a pilot study

• Include manipulation checks

review the literature

see what has already been able to show variability (differences/utility)

--> use what has already been done

pilot study

Small preliminary study used to determine the extent to which a manipulation or measure shows an effect

--> test it out yourself to see what works

Manipulation checks

• Procedure to check that a manipulation in a study had the effect that was intended

• Use multiple measures

Example of manipulation checks

two different conditions: (1) group in dark room (2) group with all the lights on

--> show both a scary movie

--> see if movie in dark has greater affect that movie in the light

--> look at heartrate to see if their are differences between conditions

--> might check heartrate before the movie (see differences with just the different lighting)

Manipulation can be seen in

clinical research studies

--> checking to make sure the things we see are accurate to what we are studying

sample

the small portion of a population that we study from the broader group that we are interested in

Goal of research is to

talk about human populations (ie. students, children, etc), but we can never address/assess the larger population completely

population

the broader group

--> we will NEVER be able to include ALL members of a population

Why we sample

• Lack of access to all individuals in a population

• Allows researchers to make inferences

inferential statistics

drawing inferences from a sample to see what might be true to the population

--> how the collected data from the sample could potentially relate to the larger population

Example of sampling

when you get a new pair of shoes, you can't go around asking everyone what they think, but we ask a few and use their opinion to make an inference about what most people would think

--> if a few say they like them, we will assume that the majority of people will like them

Participants

used for humans involved in research

• Humans only

• Provide informed consent

Subject (of research)

any thing involved in research that is not human (ie. animals)

• Nonhumans

• “Subjected” to research without consent

• Subjection to research determined by researchers and ethics review boards

May use subject when doing research with humans

ONLY when it refers to the DESIGN of the study

--> used as an adjective to describe the design

• Used to identify names of research design

• Refers to both nonhuman and human groups

--> Between-subjects design and Within-subjects design

Between-subjects design

when research is two separate groups of things

--> ie people, animals, etc

--> Group 1 and group 2

--> two different conditions

Within-subjects design

when we have a study where all of the individuals are going through both conditions

--> all participants/subjects are doing both things

We NEVER use subjects in

the methods section, but we can use it when discussing the design of the study (how people were assigned to a group)

we select samples from

the population; they broader group we are trying to access

target population

Large; Cannot sample directly

--> the group we are trying to access

Accessible population

Smaller; Can sample directly

--> part of the target population BUT we CAN access them

goal of selecting samples is to

Describe target population

--> we will never be able to access an entire population, so we collect a sample to serve as a representation of the target population

Example of accessible population

if our target population is adolescents in an urban setting, our accessible population would be students in Hartford

--> still a lot of people that we can access, but we still won't get everyone

Representative sampling

A sample that resembles the target population

example of representative sampling

someone is looking at college students, if they sample from UConn Hartford versus Storrs --> there would be differences

--> challenging to say which campus is more representative as college students as a whole, but Hartford would most likely be more representative of population at large because it has more diversity

Having a sample that is more

representative of the population allows us to say that the findings are more likely to accurately represent the larger population

Probability sampling

Sample direct from target population

--> RARE

Probability sampling is difficult

it requires the population to be split by demographic and then apply that for the sample

--> the probability of something existing in the population

Nonprobability sampling

• Sample from accessible population

• Convenience sampling

Convenience sampling

• Participants selected based on convenience

• Role of participant pools

• Drawbacks

participant pool

a group of potential research participants (usually college students or lab rats).

drawbacks of convenience sampling

sample is supposed to be generalized back to the population, but a convenience sample does not allow us to make that generalization

quota sampling

Knowing difference between some variable of interest in the population and trying to create those differences/representations in the sample

• Subjects selected based on known or unknown criteria in target population

• Ensures representation of subject characteristics

Example of simple quota sampling

splitting a sample into half men and half women;

--> the real population if split 51% women and 49% men

splitting half freshman/ half sophomores

--> not truly representative but still a quota; simply splitting into "half" even if does not truly represent the population

proportionate quota sampling

reflecting the proportions of a population in the sample

--> if we define certain percentages of the population's demographics and then mirror that in the sample

example of proportionate quota sampling

25% obese and 75% non obese in the population

--> the sample would try to reflect this proportion

we can do a proportionate quota sample

for anything that we have an identified population statistic

--> ie. men v women

simple quota sampling

cutting up into whatever percentage makes sense, not defining percentage based on what truly exists in the population, just merely breaking it apart

Proportionate quota sampling relates to

inference and reflecting back on the findings to see how they apply to the population

Simple random sampling

Must know likelihood of selecting each individual

--> relates back to quota sampling

Drawbacks of SRS

may not be representative of population

two types of SRS

1. Simple Stratified Random Sampling (SSRS)

2. Proportionate Stratified Random Sampling (PSRS)

Simple stratified random sampling

an equal number of participants is selected from each subgroup

--> may not reflect the true population

--> when subgroups contain similar numbers of persons in the population

proportionate stratified random sampling

a proportion of participants is selected from each subgroup to resemble the proportion of those in the population

--> all add up to 100%

--> may not necessarily represent the population either

--> used when subgroups are NOT equal

Systematic sampling

First participant chosen is random, but following participants can be selected through a system

• Simple random sampling on first participant

• Subsequent participants selected using systematic procedure (i.e. selecting every nth person)

Random sampling

everyone in the population has an equal chance of being selected

Example of systematic sampling

in gym when they pick one person to start off the counting to four and then as they go down every four people is in a different group (ones gets grouped, twos get grouped, and so on)

Systematic sampling is NOT

random sampling

Sampling error

Difference between observed sample and true population

--> if sample we obtain doesn't reflect the population

example of sampling error

class taught at both 9am and 10am

--> 9am did better but there are many factors that contribute to these differences

--> doing study at different times can create different results

--> selecting some people contributes to difference results

Standard error of the mean

• Distance that sample mean values can deviate from population

• Numeric measure of sampling error; numerical way of accounting for sampling error

To reduce standard error

increase the sample size

--> takes into account more of the population

Sampling bias (or selection bias)

Sampling procedures favor certain individuals or groups over others

--> people not having access creates bias

examples of sampling bias

ad only posted on an online forum will only reach the people that are on the forum

people without phones/computers cannot access online study

News and sampling bias

if news channels do surveys, they are only really getting data from their viewers (ie. liberal or conservative), so their "studies" are biased in the direction of the people that watch their channel

Nonreponse Bias

• Sampling favors participants that request to participate in a research study

• Misses participants that choose not to respond

--> missing out on everyone who chooses not to participate in the study

background of participant pools

Creation of collegiate participant pools that require students to participate in research studies

--> help students participate in research

Ethical concerns of participant pools

Justice

• Overburdening of college students to participate in research (they have homework, exams, and other things to do)

Two rules that address ethical concerns

• Class grades never contingent on participation in research study

• Alternative options given to students to receive a grade

Conducting a study is important

because it allows you to make observations using the scientific process to answer your research question

hypothesis should

lead you in the direction of this type of design answers/best serves this question

The type of study you conduct

depends on the type of question you are asking

research design

Specific methods and procedures used to answer a research question

--> Types of research questions are categorized as exploratory, descriptive, or relational

Types of research questions

- exploratory

- descriptive

- relational

Exploratory questions

"what if" questions

--> to get an idea of or explore an area of research that is not well understood

--> rarely provide definitive answers, but they lead to stronger focus for subsequent research

examples of exploratory questions

- what if a high fat/high sugar diet is physically addictive?

- what if human memory has an infinite capacity for storage?

Descriptive questions

"What is" and "How"

--> define something and describe the process

--> concerned with simply describing variables

example of descriptive questions

- what is the average time spent watching TV per year?

- how many pounds does a college student typically gain in their freshman year?

Relational questions

"does" and "is" types of questions

--> extent to which specified relationships exist between variables

--> provide (1) causal explanations or (2) descriptions of the relationship between two or more variables

examples of relational questions

- do low levels of serotonin in the brain cause depression (causal statement)

- is personal income related to life satisfaction (simply talking about a relationship not a causal statement)

Relational questions are related to

experimental and correlation designs

3 types of research designs

- experimental

- quasi-experimental

- non-experimental