Cm321 Exam #2

sampling

the science of systematically drawing a valid group of subjects from a population reliably

sampling methods

probability (random) sampling (used to generalize to a larger population) and non-probability sampling (used when you are looking for a specific characteristic)

population

the entire group of people (or messages) you are interested in learning about

census 

when the entire population is measured 

probability/random sampling

all members of the population must have an equal chance of being included and then elements of that population are then randomly selected 

simple random sampling

obtain a list of all population members, assign number to all members, randomly select numbers until desired sample size is reached; easiest way to sample but hard to get a list of the entire population

Systematic random sampling

obtain a list of all population members, assign numbers to all members, randomly select a starting point, select every “kth” element from a list

stratified random sampling

divide population into “strata” and randomly select subjects from each strata based on proportions in population; the idea is to increase the representativeness of strata 

multi-stage cluster sampling 

population is divided into “clusters”; randomly select “clusters”; randomly select participants from within the clusters

non-probability sampling

any method where a member of a population does not have an equal chance of being selected; typically used in qualitative data; in most cases this data cannot be generalized

convenience sampling

when subjects are selected based on available to the researcher; certain group may be overrepresented while others are excluded

volunteer sampling

individuals volunteer to be included but since only those willing to participate are included, the same may not be representative of the general population

purposive sampling

subjects selected for a good reason tied to purposes of research; useful for populations that are not easily obtained by screening general population 

snowball sampling

the researcher approaches one subject, asks that subject to suggest others and it continues; useful for hard to reach subjects 

quota sampling

research samples with a certain number of subjects in various categories; commonly applied in political polling and consumer research to reflect demographic distributions

survey research 

any time you systematically ask people about their attitudes, emotions, beliefs, knowledge, intentions, or behaviors

primary goals of survey research 

identify/describe perceptions, attitudes or behaviors, examine relationships between variables measured 

types of survey questions

close-ended and open-ended

close-ended survey questions

primary question types in survey; participants choose from provided responses

open-ended survey questions

primary question types in focus group or other qualitative methods; participants generate responses which allows for more details, allows for unforeseeable responses

closed-ended question types

check all that apply, rank ordering, multiple-choice, likert scales, semantic differential scales 

Likert scale

the most commonly used scale in mass comm research, this scale asks participants to respond statement with a 5 or 7 point scale

Likert and Likert-type categories

agreement (strongly disagree—> strongly agree), frequency (never to always) and satisfaction (completely satisfied to disstatisfied) 

purpose of experiments 

to establish causation between variables/events; to establish causation we need strong internal validity and are willing to sacrifice external validity 

criteria for causation

1. cause and effect must be correlated 

2. Cause must precede the effect in time 

3. There are no plausible alternative explanation

Threats to internal validity

1. History: an external event occurs that affects the results of the study 

2. Maturation: respondents change over time, which affects the DV/outcome

3. Experimenter bias: attitudes or behavior of researcher affect results 

4. testing/sensitization: participant is familiar with the measure 

5. regression to the mean: people may score higher or lower on a measure, then move toward the mean when measured again 

6. Experimental mortality: people dropping out of the study 

7. contamination: people who have been part of the study tell others what the study 

8. Sample bias/non-equivalent groups: groups in conditions are not equivalent before starting study

Key elements of an experiment

• manipulation of independent variable(s)

  • create different conditions/groups that receive different treatments

• Random assignment of participants

experimental design notations

R= random assignment 

X= manipulation/treatment

O= Observation

Factorial design

experimental studies with two or more independent variables; to understand whether the combination of two or more variables increases/decreases effects

repeated-measure design/within-subjects design

every subject completes every treatment/manipulation

within-subject designs advantages

don’t have to worry about individual differences; fewer participants are required

within-subjects designs disadvantages

order effects, fatigue of the experiment

Solution: counter-balanced design

between-subjects design

take a sample, then split the sample into groups and then give each group a different treatment

repeated-measure design/within-subjects design

take a sample, then give the entire sample one treatment, then another

cleaning data

removing “bad” cases, things like a fast completion time, incomplete data cases or straight line answers

Descriptive statistics

summarizes and describe the pattern of a variable

frequency

number of time a given response is reported

central tendency

the point in the distribution where the data are centered (how are the data similar?)

Dispersion

how spread out the data are (how are the data different?)

central tendency: mean

the mean, take everyone’s score on a variable and divide by the number of scores; must have continuous data (either ratio or interval level)

central tendency: median

the middle response when all responses are ordered from least to greatest (or the average of the middle 2 responses)

central tendency: mode

the response that occurs most often

dispersion: range

the difference between the largest and smallest observations

dispersion: deviation

how different is each individual case to the mean

dispersion: standard deviation

the dispersion of a dataset relative to its mean, aka the average deviation

what does it mean if the standard deviation is big?

the scores tend to be spread out; the average separation between the mean score and any individual’s score is large 

what does it mean if the standard deviation is small?

the scores tend to be close together; in other words, the average separation between the mean score and any individual’s score is small 

hypothesis

a specific testable statement of relationship between two variables based on theory or prior research; in order to test our hypotheses we need to use inferential statistics 

Null hypothesis (H0)

states that there is not a relationship between two variables; think of it as the “devil’s advocate”; we must start with the assumption that no relationship exists and find evidence to the contrary; the default is the null is valid 

alternative hypothesis (Ha or H1)

there is a difference/relationship/influence; we want Ha to be true; think of H0 as the status quo, and Ha as change or innovation; we cannot directly say we found evidence to support Ha instead we must say we found evidence to reject the null hypothesis; hypothesis testing tells us if we reject or fail to reject the null hypothesis 

directional hypothesis

something causes more or less of another

Non-directional hypothesis

there is a difference between 2 things

Probability and hypothesis testing

whether or not we reject or fail to reject the null hypothesis depends on the probability that we are obtaining our results by chance 

P value or significance value

Low p: reject null (there is a difference)

High p: fail to reject null (no difference found)

Most common cutoff for p value:

5% chance of being wrong (p<0.05), if a p value is less than 0.05, you can reject the null or 1% chance of being wrong (p<0.01), if a p value is less than 0.01, you can also reject the null

type 1 error

false positive; wrongly reject a H0 when it’s actually true; finding a difference that’s false

type 2 error

false negative; fail to reject H0 when it’s actually false; not finding a difference that is actually there

Categorical levels of measurement

nominal and ordinal

Continuous or quantitative levels of measurement

interval and ratio

statistical program used: categorical dependent variable and categorical independent variable

cross-tabs

statistical program used: quantitative dependent variable and categorical independent variable

ANOVA

statistical program used: categorical dependent variable and quantitative independent variable

Rarely happens

statistical program used: quantitative dependent variable and categorical independent variable

Correlation

correlation

measures if there is a relationship between two variables