PSYC317 - EXAM 2

sample

a subset of individuals in the population

sampling

the process by which a researcher selects a sample of participants for a study

what would you ideally want when sampling

ideally, you want your sample to be representative of the population

probability samples

a sample is selected in such a way that the researcher can estimate an individual member of the population’s statistical likelihood of being included in the sample

why are probability samples essential

essential for describing the behavior, thoughts, or feelings of a particular population

• need it to be as representative of the population as possible

sampling error

characteristics of the sample will always differ somewhat from the characteristics of the general population

what does sampling error cause

causes the results obtained from a sample to differ from what would have been obtained if the entire population had been studied

error of estimation (aka margin of error)

the degree to which the data abstained from the sample are expected to deviate from the population as a whole

• we can estimate the degree to which the results were affected by sampling error

what is error of estimation affected by

• sample size

• population size

• variance

how does sample size affect error of estimation

are larger or smaller samples more likely to be representative of the population?

• bigger sample = less sampling error (and smaller error of estimation)

how does population size affect error of estimation

is it easier to represent a larger population or a smaller population?

• larger population = greater sampling error (and larger error of estimation)

how does variance affect error of estimation

is it easier to represent a population where everyone is very similar, or where everyone is different?

• greater variance = greater sampling error (and larger error of estimation)

types of probability samples

• simple random sampling

• stratified random sampling

• systematic sampling

• cluster sampling

  • multistage cluster sampling

simple random sample

random = every possible sample of the desired size has the same change of being selected from the population

• this also means that every individual in the population has an equal chance of being selected for the sample

when is simple random sampling essential

when the goal is to accurately describe the behavior of a particular population

what does simple random sampling require

a sampling frame: a list of the population from which the sample will be drawn

• pick names out of a hat, assign numbers to cases and use a random number generator, random digit dialing (landlines only)

why are simple random samples rare in psychology research

• often, researchers do not have a sampling frame (a list of every person in the population)

• expensive, time-consuming

• generally, not necessary

  • rarely are we trying to describe the behavior of an entire population

  • instead, looking at relationships between variables

  • can replicate on other samples to see if it generalizes to a larger population

stratified random sampling

• divide the population into two or more subgroups or strata

• then, randomly select participants from each stratum

stratum

a subset of the population that shares a particular characteristic

is stratified random sampling still random

yes, everyone still has an equal chance of being included

does stratified random sampling need a sampling frame

yes

proportionate sampling method

cases are sampled from each stratum in proportion to their prevalence in the population

example of proportionate sampling method

let’s say we want our sample to be proportionate to the racial makeup of VCU

• VCU has 40% white students, so we would gather a list of all White students and randomly select 40 white students

• VCU has 21% Black/ African American students, so we would gather a list of all Black/ African American students and randomly select 21 Black students

• ets. for all racial/ ethnic groups

does proportionate sampling method count as a probability sample

yes, it is able to estimate probabilities

does proportionate sampling method count as a random sample

yes, everyone still has an equal chance to be selected

cluster sampling

randomly select naturally-occurring groupings/ clusters of participants

what is the issue with doing a simple random sample when we want to interview students in public, four-year colleges/ universities in the state of virginia

researchers would end up driving to many campuses, some maybe for one participant

multistage cluster sampling

randomly sample large clusters, then successively smaller clusters within the large cluster, until finally obtaining the sample

cluster sampling example

• first, I would need to get a sampling frame of all public, four-year colleges/ universities in Virginia

• then, I would randomly select clusters; here, the universities

  • I might randomly choose 5 of the schools

• depending on how big the clusters are, I could sample the entire cluster

  • since a cluster would be an entire college in this example, I would probably need to randomly sample students within each of the 5 chosen schools

example of multistage cluster sampling

randomly select 5 → randomly select 3 dorms from each of these universities → randomly select 25 students from each of the 3 dorms

systematic sampling

taking every so many individuals for the sample

is systematic sampling random

NO!! after a participant is selected, the next several people have no chance of being in the sample

non-response problem

• the people that we select for our sample do not always respond or agree to participate

  • this could be due to lack of time, inconvenience, disinterest, distrust of the researcher, sense of being used without appropriate compensation, fear of private info being leaked, etc.

• when people who were randomly selected do not respond, this biased our sample

  • people who don’t respond may be different in some way from those that do, making the sample no longer representative

misgeneralization

generalizing results from a study to a population that differs in important ways from the one from which the sample was drawn

example of misgeneralization

you read a headline that says “study finds that Virginia college students are no longer in need of financial aid”

• like a good PSYC317 student, you look at the original article, and see that they only randomly selected students from private universities, no public

nonprobability samples

samples in which the researcher cannot estimate the probability that a particular case will be chosen for the sample

what can nonprobability samples not estimate

cannot estimate error of estimation

what are the two types of nonprobability samples

• convenience samples

• quota samples

convenience samples

includes whichever participants are readily available

example of convenience samples

studying adolescent depression and sleep

• instead of trying to sample all adolescents across the country, sample from local pediatric offices

is convenience samples bad science

NO

• it’s not trying to describe a population, it’s looking at relationship between variables

• can replicate with other adolescent samples to see if the findings generalize to other groups

historically, the vast majority of psychology research (67% of American research) has been conducted on college students

TRUE

limitations of convenience samples

• findings are often generalized to the entire adult population (US or even globally)

• college students may differ from the general population in important ways

• college students are mostly from WEIRD (white, educated, industrialized, rich, democratic) societies

nearly 70% of studies use samples from only 12% of the world’s population

TRUE

quota samples

a convenience sample in which the researcher takes steps to ensure that certain kinds of participants are obtained in particular proportions

example of quota samples

we know that alcohol use disorder (AUD) is more prevalent in men than women

• let’s say we are doing an online survey of AUD, with a desired sample size of 100

  • we might first screen for gender

  • once we obtain 33 women participants, we might stop allowing women into the study, so that the majority of the sample will be men

are larger samples better

generally, but not always feasible

power

the ability of a research design to detect statistical effects of the variables being studied

do larger samples have more power

yes, and therefore more ability to detect an effect, especially if it’s small

descriptive research

describing the characteristics or behaviors of a sample or population

what does descriptive research not test

it’s not testing hypotheses, it’s obtaining basic information

examples of descriptive research

• 50% of our class believes that pineapples belongs on pizza

• students in our class have a mean extraversion score of 4.97 on a scale from 1-10

demographic research

describing basic life events and experiences

• birth rates, divorce rates, employment, migration, death

current population survey

• administered MONTHLY

• probability selected sample- stratified cluster

• about 60,000 occupied households

• source of national unemployment rate and other national demographic data

epidemiological research

describing death and disease in a population

• prevalence

• incidence

prevalence

the proportion of a population that has a particular disease or disorder at a point in time

incidence

the number of new cases that occur over a specified period of time

survey research

can refer to either questionnaires or interviews

• different designs based on the timing of survey administration

cross-sectional research

a single group of respondents- a “cross-section” of the population- is surveyed at one point in time

• because everyone is sampled at one time, it is cross-sectional

successive independent samples

two or more samples of respondents answer the same questions at different points in time

how is successive independent samples helpful

helpful to track changes in time, but only if the two samples are comparable

example of successive independent samples

the World Happiness Report asks people across the world the same question each year

• “taking all things together, would you say you are: Very happy, Rather happy, Not very happy, Not happy at all”

• comparing percent of happiness based on percent who answered either Very or Rather and compare happiness trends over time

longitudinal study

follows a single sample (i.e., the same participants) over time and surveys them multiple times

examples of longitudinal study

• spit for science asks the same students to provide data each year that they are in college (and a few years after)

• the Midlife in the US study (MIDUS) asks the same group of adults to provide data every few years

what is a major challenge with longitudinal research

attrition

attrition

people dropping out of the study

• are the changes that we find real, or due to the sample changing because some have dropped out

advantages of internet surveys

• cost efficient

• convenient for both participant and researcher

• expand recruitment boundary (i.e., reach more people)

• reduces data entry errors

disadvantages of internet surveys

• less control over sample characteristics

• bots

• possible fraud

• not every has access to internet

• overrepresentation of low income, less educated, over 65

correlation research

investigates the relationship between two or more variables

• looking at whether two variables covary- do they vary or change together

• change = increase or decrease

example of correlational research

• is self-esteem related to shyness

• what is the relationship between income and happiness

• is gender associated with depression

can correlational studies determine causation

NO

how do we describe the relationships between two variables

• visually with a scatterplot

• statistically with a correlation

both are merely descriptive, and no causation can be inferred

scatterplot

each case (person) has a score on both variables

• the more tightly the data points are clustered around an imaginary line running through them = the stronger the correlation

correlation coefficient

a statistic that indicates the degree to which two variables are related to one another in a linear fashion, or the degree of co-variation

what is the correlation coefficient denoted by

r

what does the type of correlation we use depend on

depends on the scale of measurement of our variables

• most common: pearson correlation coefficient, used if both variables are interval/ ratio

pearson correlation coefficient

• ranges from -1.00 to +1.00

• tells you

  • the direction of the relationship

  • the magnitude of the relationship

positive correlation

as one variable increases, the other variable increases

• correlation coefficient would be a positive number

negative correlation

as one variable increases, the other variable decreases

• correlation coefficient would be a negative number

we can tell how strongly the two variables are associated by…

• how tightly the points are clustered together in a scatterplot

• how big the numerical value of the correlation coefficient is, ignoring the sign

correlation coefficient with an absolute value below .29 are considered…

weak

correlation coefficient with an absolute value from .30-.49 is considered…

moderate

correlation coefficient with an absolute value from .50-1.00 is considered…

strong

does a correlation coefficient of .00 always mean no relationship

sometimes, two variables have a curvilinear relationship

• the correlation coefficient would be very low (.00), but technically there is a relationship there

is a coefficient of .80 twice as large as .40

NO, r is not on a ratio scale, and not directly interpretable

• must square r to create a ratio scale r²

coefficient of determination

r²

• explains how much variation in Y is explained by X

• same a effect size

  • go beyond statistical significance to explore the STRENGTH of that relationship

coefficient of determination example

how much of the variation in depression (Y) can be explained by sleep quality (X)?

• Y may be affected by many possible factors, but we are interested in the unique contribution of X

• r = -.64

  • coefficient of determination = -.64² = .41

• thus, we can say that 41% of the variability in depression is attributable to differences in sleep quality

statistical significance

a correlation coefficient calculated on a sample has a very low probability of being zero in the population

• p < .05