CMM 412 Unit 3 Study Guide

dyadic/triadic contrast

present two/three options of specific characteristics

descriptive statistics

statistical procedures used to *summarize*, *organize* and *simplify* data

Step 3: Select the Appropriate Statistical Test

when should we use independent sample t? - *Nominal IV?*: Male and female? - *Interval/Ratio DV?*: Comm apprehension scale from 24 to 120? - *Samples are independent?* Scores for males don't affect scores for females & vice versa?

Step 9: Write up your analysis (in APA)

Know how to do this! Name statistical test Name variables Both strength and +/- r, df, and p values Explain in words

4. Deciding to reject or fail to reject the null hypothesis

Make a Decision: - *Compare* the *calculated* statistic value to the *critical* value - *Critical region*: unlikely sample values - Does our calculated statistic fall into the *critical region*? If NO, *fail to reject* the null hypothesis. If YES, *reject* the null hypothesis. Do not use the word ACCEPT.

frequency distributions

Organized *tabulation or picture* of the number of data points

central tendency: mean

*"Average"* of all the scores *Procedure:* - Add up all the scores - Divide by the total number of scores Example of calculating the mean: 8 scores: 1, 2, 1, 2, 4, 2, 3, 5; add up = 20; 20/8= mean of 2.5 *Problem with Mean?* - Means are subject to *outliers*, or numbers in the set that are *far away* from the mean E.g. home prices; income - The *median* corrects for this shortcoming

trustworthiness of interviewing

*Dependability*: can we track, take notes, record? *Confirmability*: conclusions warranted by evidence? *Credibility*: "ring true" to respondents? Can verify with more interviews... *Transferability*: cannot make generalizable claims, but can rich enough picture allow you to see relevance?

measures of central tendency

- Statistical procedure to determine a single score that defines the *center* of the distribution - *Goal*: find the single score that is most *typical* or most *representative* of the entire group - *Types*: mean, median and mode

sampling error

- The discrepancy (amount of error) between a sample statistic and its corresponding population parameter - Example: Self-esteem scores of 125 students vs. All students - *The trick*: we cannot know sampling error unless we know the population parameter

Why do we employ qualitative methods?

- We use qualitative methods to get *rich detail* (not generalizability) - Participant observation (e.g., ethnography) - *Interviews*: conversations - *Focus groups* (important in campaigns!)

variability: standard deviation

- What we use *most* to describe variability - Tells us whether the scores are *generally*: Near the mean (low standard deviation) or Generally far from the mean (high standard deviation) - On average, how far are the scores from the mean? - *Standardized* way of looking at the differences - *68.26%* of a sample always falls within *one* standard deviation of the mean - *95.44%* of a sample always falls within *two* standard deviations of the mean

correlations

- assess *linear relationships* (have no changes in direction; vs. *curvilinear*: more than one changes in direction) - e.g., "As X increases by one unit, Y increases or decreases by one unit." - *IVs & DVs?:* Both interval or ratio (scale) - Both variables should be *independent*

variability: variance

- measures *how widely spread* out a distribution is: the average distance of each score from the mean - as variance increases, it shows *greater variability* in the data

variability: range

- number of value classes between the highest and lowest observed values *including* those values - it's the difference between the highest and lowest plus one 1, 2, 3, 4, 5; range = *5-1* +1 = 5 - range tells us how far apart our largest and smallest scores are, but it *DOES NOT* give us any insight into how much those scores differ from the average score

alternative hypothesis (HA)

- predicts that there *IS* a relationship or difference that is not due to chance - something that you are proposing; you test your hypothesis against a standard- the null

4 reasons for qualitative interviewing

1. To learn what we *cannot directly observe* 2. To understand experiences in richly *detailed* manner Resist *subjectivity* Understand *bias* cannot be removed 3. To understand informants' *language* - vocabulary and idioms 4. For *triangulation*: compare findings to other methods

central tendency: median

Exact *middle* value *Procedure:* - List all the data in order from smallest to largest - If *odd* number of values: number at the exact midpoint - If *even* number of values: add the two midpoint numbers and divide by two

1. State Alternative and Null Hypotheses

HA: Men and women *differ* in their communication apprehension. H0: Men and women *do not differ* in their comm apprehension.

Step 1: State Alternative and Null Hypotheses

HA: There is a *relationship* between levels of communication apprehension and change in heart rate during a five-minute impromptu speech. H0: There is *no relationship* between levels of communication apprehension and change in heart rate during a five-minute impromptu speech.

HA and H0 examples

HA: There is a difference between men and women in terms of verbal aggression (two-tailed). H0: There is no difference between men and women in terms of verbal aggression. HA: There is a relationship between outgoingness and attractiveness. H0: There is NO relationship between outgoingness and attractiveness.

other types of t-tests: one-sample

Is the *sample* mean significantly different from a known *population* mean?

Step 9: Write up your analysis (in APA)

Know how to do this!

How do I know *which* measure of central tendency to use?

Often use more than one measure to describe a sample

central tendency: mode

Value that occurs *most frequently* *Procedure:* - Count how many *times each value occurs* (easiest if they are in order) - The one that *occurs the most* is the mode - Note: if *two or more values* occur in the same number of times, they are *all modes*

skewed distributions

Values pile up toward one end of the scale ("PEAK") and taper off gradually at the other end ("TAIL") POSITIVELY SKEWED - Peak of curve (mode) is on the *left* - The "tail" is on the *right* - *Median*: *right* of the mode (i.e., pulled in *positive* direction) - *Mean*: *right* of the mode and median (i.e., pulled in *positive* direction) - *Mean > Median or Mode* NEGATIVELY SKEWED - Peak of curve (mode) is on the *right* - The "tail" is on the *left* - *Median*: *left* of the mode (i.e., pulled in *negative* direction) - *Mean*: *left* of the mode and median (i.e., pulled in *negative* direction) - *Mean < Median or Mode*

Step 3: Select the appropriate statistical test

When to use a Pearson product moment correlation? (r statistic) IV? DV? Alternative: Spearman's Rho = rank order correlation, non-parametric

interview protocol: structured

each respondent gets *same* set of questions common of surveys, limited in interviews...

1. DESCRIPTIVE QUESTIONS

own words, rich detail *Grand tour (typical, specific, guided, task-related)* *Example* *Experience* *Natural language*

parameter vs. statistics

parameter: *population* statistics: *sample*

null hypothesis (H0)

predicts that there is *NO* relationship or difference (it is NOT just the opposite effect)

hypothesis testing

process of using *inferential statistics* to determine whether we *reject* or *fail to reject* the null hypothesis

ratings

rate what is "best," "worst," "most," etc.

included-term

start with who/what, ask to identify group (ex: this person seems... must be an x major)

inferential statistics

statistical techniques that allow us to *make generalizations* about the populations from which the samples were selected E.g., Hypothesis Testing - "infer" = generalizing from data that we have - making claims of *difference* and of *relationship*

substitution frame

take a statement, ask respondent to complete

grand tour

verbal "map" of phenomenon *Typical*: verbal map of what is "typical" (ex: turkey, family ~ Thanksgiving) *Specific*: recent day, event, activity, etc. (ex: "What did you do last Thanksgiving?" "What are you doing this Thanksgiving?") *Guided*: physical tour (ex: describe the setting, orient the space; what's the process- what it feels like/ looks like) *Task related*: do task, describe (ex: family fitness class Thanksgiving morning)

natural language

what are some of the terms (lingo) used, related to a culture

standard error

- Standard deviation of our sample divided by the square root of our sample size - As our *sample size increases*, our *standard error decreases*

genres of interviews

*Ethnographic conversation*: Informal, emerge from interaction Hard to separate from participant observation *Depth*: longer, formalized; stand-alone *Group interview*: two or more people Added insight by observing *interaction* *Focus group* (7-12 people) - Typically need to conduct multiple focus groups - Relevant participants selected (not probability) - Homogenous within group - Role of moderator important (want them to blend in with group) *Narrative interview*: use stories to make sense - *Life history*: link individual's life to societal, cultural themes - *Oral history*: gather stories about individual's life or several individuals for period of time *Postmodern interview*: Critical theory driven Interviewer empowers, gives voice

types of correlations

*Positive relationship*: variables move in the *same* direction *Negative/Inverse relationship*: variables move in the *opposite* direction *No Relationship*: change in one variable --> *anything* can happen Strength of relationships: - The closer points are to the line --> stronger the relationship

interview protocol: semi-structured

*general* list of questions can vary wording, paraphrase to be more conversational

sampling distributions

- *Frequency* that values of statistics are observed or expected to be observed when a *several* random samples are drawn from a given population - Is the distribution of a statistic across an infinite number of samples - Take many *different random samples* from the population - Take the *mean* (a statistic) of each of the samples - Frequency that each mean occurs can be drawn in a frequency distribution → the sampling distribution

building rapport

- *Rapport*: harmony, accordance, affinity - Focus on content of interview *AND* interaction - View as a *dialogue* - *Trust* and *respect* - Be *transparent* - Do *"research"* before

crosstabulations

- A way of presenting, reporting data - Summarizing data to compare between groups

chi-square test

- Could use to make statistical comparisons between frequencies or groups (we will not calculate in this class) - Say we're interested in people's favorite music. We ask people from 3 diff geographical regions about their favorite type of music.

independent sample t-test

- Determine if there is a significant *difference in the mean* score on DV between *two groups* (IV) - "Independent" because membership in one group does not influence membership in the other (male vs. female? NOT married couples) - *Variables for independent samples t-test*: *IV*: Nominal variable with *TWO* LEVELS (e.g., male vs. female; *NOT* Democrat, Republican, Independent; *NOT* Freshman, Sophomore, Junior, Senior) *DV*: Interval or ratio variable Examples: - Does gender influence how much one likes video games? - Does hallmark or freeform have more engaging Christmas movies? - Who are considered "cooler": Android or iPhone phone users?

use *mean* when:

- Interval or ratio measure - Normal distribution

other types of t-tests: paired sample

- Is there a difference between sample means *measured twice*? - E.g., Time 1 and Time 2 (like pre- and post-test in an experiment)

doing Q&A's

- Let informants use *own words*, pace - *Probe* for more details - Transition in ways that *guide*, not control - *Paraphrase* - restate what was said - Display *supportiveness* - Be *flexible* and adaptive

variability: kurtosis

- Measure of the *"peakedness"* - Sense of the variability of the curve - *Platykurtic:* flat distribution (like a plate) - *Leptokurtic:* peaked distribution (like the cursive letter "L")

use *mode* when:

- Nominal variable - Describe the shape

parameters

- Numeric value that describes the *population* - Need info from *EVERY SINGLE* member of population - E.g., *all* Fall 2019 UD undergrads enrolled Women: 3,875 (48%) Men: 4,171 (52%)

statistics

- Numeric values that describe the *sample* you've collected (when we do a study, we collect a *sample*, a *statistical* test) - E.g., *sample* of 125 Fall 2019 undergrads enrolled Women: 40 (48%) Men: 65 (52%) - Statistical test reports the probability the *parameter* falls within the confidence interval - Degree of uncertainty because of *sampling error*

use *median* when:

- Ordinal measure - Extreme scores or skewed distributions

1. Establishing Type I and Type II Error Risk levels

-Two Types of Sample Means: - Sample means that are *likely* to be obtained if the null hypothesis is true (sample means that are *very close* to the null hypothesis) - Sample means that are *very unlikely* to be obtained in the null hypothesis is true (sample means that are very *different* from the null hypothesis) Significance level/p-value/alpha level: - Probability level used to define the *very unlikely* sample outcomes if the null hypothesis is true - *Typically choose a = .05* - Meaning: if null is true, we expect to see the unlikely sample means 5% of the time *due to chance* - Significance level reported in SPSS as *Sig. Level* or *p* - Tells us if our sample has probably occurred by chance or not p > 0.05 → it's just chance (likely values) p < 0.05 → our differences are likely *NOT* just chance - *Confidence level* is related to significance level 95% confidence level → 0.05 significance level 1 - .05 = 95% confidence level Power: - ability statistically to find a relationship or difference in your *sample* if it exists in the *population* - The *greater* your sample size, the *greater* your power - Smaller the difference → more power you need (e.g. microscope analogy) *Type I Error*: - Researcher rejects the null hypothesis when the null hypothesis is true - Researcher concludes that there *IS* an effect or difference when in fact there *IS NOT* - *False positive* - *a* is likelihood of Type I error ~ relates to *confidence* *Type II Error*: - Researcher fails to reject the null hypothesis when it is false - Researcher concludes that there *IS NO* effect or difference when in fact there *IS* - *False negative* - *Beta* (1 - *power*) is the likelihood of Type II error

Step 4: Determine the Critical Value (needed for rejection of the null hypothesis)

1. Determine your *degrees of freedom* (df) 2. Use independent samples t table to determine critical value for t statistic (at your α level and your df) Calculate degrees of freedom: (df) = n1 + n2 - 2 n1 = number of units (people) in the 1st level of the IV (# of men in our study) n2 = number of units in the 2nd level of the IV (# of women in our study) - Use an independent sample t-test critical value table - We calculate out *t-value* to compare to *critical value* - t = 0 if there is no difference between the two groups - t can be positive or negative - Difference —> absolute value of t is —> the more likely you are to have a significant result - If absolute value of the calculated t (|t|) is *greater or equal to* the critical value (1.96), we *reject* the null hypothesis - If *less* than, we *fail to reject*

steps in hypothesis testing

1. Establishing Type I and Type II Error 2. Selecting the appropriate test/3. Consulting the appropriate statistical table (AKA determining the critical value) 4. Deciding to reject or not reject the null hypothesis

three assumptions of inferential statistics

1. Sample must be drawn from population about which inferences are being made 2. Simple random sampling 3. Non-sampling errors introduced Essential: *Confidence level* *Confidence interval*

Steps to Conducting Pearson Product Moment Correlations

1. State Alternative and Null Hypotheses 2. Set the significance level 3. Select the appropriate statistical test 4. Determine the critical value 5. Compute the sample statistic 6. Compare statistic to critical value 7. Decide to reject or fail to reject null hypothesis 8. Determining meaning for your alternative hypothesis 9. Write up the analysis in APA style

Steps to Conducting and Independent Sample T-Test

1. State Alternative and Null Hypotheses 2. Set the significance level 3. Select the appropriate statistical test 4. Determine the critical value 5. Compute the sample statistic 6. Compare statistic to critical value 7. Decide to reject or fail to reject null hypothesis 8. Determining meaning for your alternative hypothesis 9. Write up the analysis in APA style

Step 5: Compute the sample statistic

Assumptions for r we check *before we conduct test*: - Variables are *interval/ratio* - Relationship is *linear* - Variables are *independent* Assumptions for r we check *after data collection*: - *Linear* relationship - *Normal* distribution for both variables (N > 25) In SPSS: * = significant at p < .05 ** = significant at p < .01

Step 5: Compute the Sample Statistic

Assumptions for t we check *before we conduct test*: - IV is nominal - DV is interval - Sample is independent Assumptions for t we check *after data collection*: - Dependent variable has *normal* distribution for both levels of the IV - *Homogeneity of variance*: samples are drawn from populations that have equal variances for the DV ----- *Levene's Test for Equality of Variances* {how we figure out which stat to LOOK AT} - Significant results (p < .05) means there *IS* a difference in the variances of the two groups - Means *violates* assumption - Use the info on the *"equal variances not assumed"* line

Step 4: Determine the critical value (needed for rejection of the null hypothesis)

Calculate *degrees of freedom (df)*: For correlation: N - 2 N # of people *who answered BOTH variables* Use *r table* to determine critical value for r statistic (at your α level and your degrees of freedom); we look at *two-tailed* *r ranges from -1 to 1* The *closer* r is to 1 or -1, the more *strongly* the variables are related r = 0 means *no relationship* r = 1 or -1 if there is a *perfect relationship* Relationship --> absolute value of r will be bigger --> the more likely a significant result Unlike with a t-test, the +/- is *meaningful*! r is *positive* for a *positive* relationship r is *negative* for a *negative* or inverse relationship

symmetrical distributions

Can draw a line through the middle, same image on either side

Step 2: Set the Significance Level (α)

Choose *α = .05*

Step 2: Set the Significance Level (a)

Choose a = .05 5% or .05 chance of making a Type I Error 95% or .95 confidence (1-a)

2. Selecting the appropriate test/3. Consulting the appropriate statistical table (AKA determining the critical value)

Collect data, use the *appropriate test* to your H or RQ: *T-test*: statistic of differences between two groups - IV: binary (two categories) - DV: interval or ratio - *Gives us t value + p-value* (to decide whether mean is likely or unlikely value) (SPSS will provide table to get p-value from test statistic) Ex: HA: There are differences btw college students and non-college students on perceptions of study drugs. T-TEST *Correlation*: statistic of relationship between two variables - IV: interval or ratio - DV: interval or ratio - *Gives us r value + p-value* Ex: HA: There is a relationship between cyberbullying and self-esteem. CORRELATION HA: There is a relationship btw social media use and anxiety. CORRELATION

Step 8: Determine meaning for your alternative hypothesis

Correlation does NOT mean causation! Determine *positive or negative* relationship: - Look at whether r is +/- Practical significance - guidelines for relationship strength: *Weak* and questionable: |r| < .30 *Moderate* and clear: .30 < |r| < .60 *Strong* and clear: |r| > .60

variability

Quantitative measure of the degree to which scores in a distribution are *spread out* or *clustered* together Measures of Variability: - Kurtosis - Range - Variance - Standard Deviation

interview protocol: unstructured

areas or *talking points* create questions in moment

contrast verification

ask respondent to confirm/disconfirm two phenomenon are similar/different

example

ask respondent to generate one

experience

ask respondent to think of one

directed contrast

ask what differentiates groups (ex: comm management vs. PR?)

cover term

ask who/what belongs in this group (ex: stereotypical high school groups in The Breakfast Club)

2. STRUCTURAL QUESTIONS

get a sense of systems in place *Cover term* *Included-term* *Substitution frame*

3. CONTRAST QUESTIONS

giving people options to probe them *Contrast verification* *Directed contrast* *Ratings* *Dyadic/triadic contrast*