Conducting Research - Tusculum University



Conducting Research

by Lawrence T. Orcher

This text is for those doing their first empirical study

Chapters 1-10 go over setting up to prepare a proposal

Model literature reviews through basic stats are included in this book

Chapter 1

Selecting Tentative Topic

Research is process of systematically collecting and interpreting information

Library research – collect/interpret info others wrote

Empirical research – making observations to collect new info (data) which is analyzed to assist in making interpretations

Observations can be overt (direct) or indirect (interviews, surveys)

Chapter 1 (cont.)

Several broad areas of interest identified based on

Everyday observations

Theories (test a theory)

Beware of availability of participants, ethical considerations (no phys/psych harm), audience (my approval), personal needs (your interests)

Can replicate published studies

Use demographics to narrow OR broaden a topic

Start searching the literature after broad area identified then may narrow or even change topic

Chapter 2 - Locating Literature and Refining the Topic

Desire SCHOLARLY/PEER REVIEWED/REFEREED articles (most recent and are checked)

Databases of published research available through paid subscriptions (by TC)

ERIC; Infotrac I – Onefile; Proquest–Education Journals

Use the database thesaurus to better define topic and find related terms

Topic should narrow, become better defined after literature search is completed

Chapter 2 (cont.)

Goal is 50-100 abstracts to examine at one time

Use Boolean operators to narrow (AND) or widen (OR) search

Review both negative and positive studies

Focus on recent studies (last 10 years)

Tutorial on APA citations on TC webpage, select Library Homepage, then Tutorials

Good idea to print the APA Quick Reference Guide

For federal statistics, searching by topic, go to

Chapter 3–Preparing (writing) a Literature Review

Lit review establishes need for your study and gives context to it

See table in text for organizing sources

Lit review is an essay about the literature

Synthesize studies about a topic

Do a topic outline first (see example in text)

Group sources together under given topic

NOT a source-by-source listing/discussion

Use APA Level 4 subheadings if lit review is long

Review examples in book

Chapter 3 (cont.)

Critical review is needed…if there is a flaw in a study, point it out (see book examples of before/after)

Avoid statements of fact/truth/proof…all research is flawed and so use tentative language (Research implies, suggests, etc.)

Point out gaps in literature, and how your study will address the gaps

Chapter 3 (cont.)

Placement of surnames for in-text citations

Use Harvard method (surnames, year)

Author can be subject of sentence (draws attention to who the author is)

Author can be parenthetical (putting the source in parenthesis puts focus on the content of the sentence…this is preferred)

Discuss relevant theories in detail before studies

Reader should see a connection between lit. rev. and RQ/purpose/hypotheses…logically RQs come at end

See 3 model lit reviews at end of book

Chapter 4 - Writing Research Hypotheses, Purposes, and Questions

These guide your study and identify the variables included

They are based on previous research or theories and logically follow the literature review

Can be given in one of three formats

Hypothesis – a statement of a predicted outcome from the study

If cause/effect the independent and dependent variables must be included

Do not give numerical values (i.e. Group X will score 50% higher than Group Y…this means you’d be wrong if it was 49% or less OR if 51% plus)

Should not be a value judgment…must be something observable

Purpose – a declarative statement that identifies the variables of interest, what was studied (no expected outcome is given)

Often this is used instead of a statement of the problem

Question (RQ) – alternate wording of the purpose, given at the end of the lit review (numbered as RQ1, RQ2…same for hypotheses H1, H2…)

Chapter 4 (cont.)

Clarifying the Purpose/Hypotheses/Questions

Often include information about the population, but not usually about the methods of measurement

Sometimes have research questions and hypotheses that link to the research purpose statement

Reviewing the literature helps clarify the RQ etc.

See text examples, especially before/after cases

Chapter 5 - Selecting a Research Approach

Choose a design (approach) after form RQ/Hyp

Experimental v. Non-Experimental

Experimental explores cause and effect relationships

Has at least one independent variable (IV)/cause which is under control of the researcher

Has at least one dependent variable (DV) or outcome/effect

Generally form two groups, one (experimental or treatment group) gets the new treatment and the other (control group) does not

Random assignment to trtmt/control key element of true experimental designs

Quasi-experimental is used when random assignment is not possible (see Ch. 22)

Chapter 5 (cont.)

Major types on non-experimental research

Causal-Comparative (aka ex post facto)

Used when random assignment to trtmt/control is not possible

Compare existing groups to identify causal sequence

Often matching is used to ensure the comparison groups are equivalent on variables that might affect the DV

Chapter 5 (cont.)

Major types on non-experimental research (continued)

Surveys (provide a description of what exists, aka descriptive research)

Survey a population (census) or a sample on attitudes/opinions

Mailed questionnaires often used (see App. A for ways to improve response rates…use incentives, keep it short, preaddressed/stamped envelopes, standard vocab/grammar)

Interviews (more labor intensive/expensive, but get more in-depth data)

Achievement surveys – measure knowledge levels

Groups’ opinions are often compared based on demographic factors (such as age, gender, race, etc.)

Chapter 5 (cont.)

Major types on non-experimental research (continued)

Correlational Studies (Appendix B has hints)

Use a correlation coefficient statistic to assess relatedness of two quantitative variables (ACT and college GPA)

Coefficient is a numerical measure from 0 to 1 (plus or minus) that tells how related the two variables are

Closer to zero, less relationship; further from zero more

Sign of the coefficient tells if the relationship is positive/direct (hi/hi and lo/lo) or negative/inverse (hi/lo and lo/hi)

Chapter 5 (cont.)

Major types on non-experimental research (continued)

Document/Content Analysis (examine existing records to look for patterns/themes)

Good samples available

Documents are secondary sources…may not be accurate

Program Evaluation (hybrid of experimental & non)

Program is a treatment, but usually not random assgnmt

Sometimes focus on implementation process (formative) rather than outcome (summative) – App. D info

Long timeframe needed

Chapter 5 (cont.)

Quantitative v. Qualitative (or breadth v. depth)

Quantitative

Numbers based, standardized procedures to obtain data, rather rigid with no personalized interactions with subjects

Generalization is important

See example list of projects in text

Qualitative

Not structured nor laid out in advance

Direct involvement with participants increases chances of bias, so usually best done by trained researchers

Hints on which to use: Some topics are inherently quantitative (including those w/ hypothesis since need structured study & measuremt); some topics can be studied with either approach; emerging topics may need qualitative

Chapter 6 - Looking Ahead to Participant Selection

How many and how selected

Size

Smaller numbers often allowed in research class projects

Experimental and qualitative usually smaller samples than quantitative and descriptive studies

Either way a minimum of 30 and maximum of about 1000

Selection method (most important…representation is essential)

Sampling methods are common criticisms of research studies

Chapter 6 cont.

Random sampling (for quantitative research) – all members of the population must be identified and have an equal chance of being selected

A random sample is an UNBIASED sample

Use names in a hat (or equivalent w/ table of random #s)

Often hard to identify all population members

Sometimes those selected do not participate (low response rates on mailed questionnaires in particular)

Stratified sampling used to ensure subgroups are represented (usually proportional…using volunteers w/in strata not ok)

Chapter 6 (cont.)

Sampling Methods (continued)

Convenience sampling (aka accidental sampling)

Presumed to be biased; using volunteers is a bad idea

Often used to pilot test instruments/materials or for exploratory studies

Findings open to criticism (should be interpreted with caution)

Purposive sampling

Often used in qualitative to gain depth in understanding a group that has information of interest (such as women officers in large corporations)

Criteria for inclusion must be clearly identified and based on specific reasons; then select a clear method to identify participants…do not use convenience OR volunteers

Chapter 6 (cont.)

Using the literature to plan sampling

Note how many tend to be used in studies reviewed in the literature and how they were selected

Pay attention to the Limitations sections of published research reports…avoid the sampling errors identified there; also can identify needed populations to study

Ch. 11/12 will cover this in more detail

Chapter 7 - Looking Ahead to Instrumentation

Methods of measurement (tests, questionnaires, interview schedules, observation forms, etc.)

Can use ones from the literature

Have established validity

Allows comparisons between studies

Any errors in them will apply to your study

Can modify ones from previous studies

Devising new instruments

Must be able to defend validity/reliability

Pilot test these

Chapter 7 (cont.)

Validity – extent to which an instrument measures what it is supposed to (Ch. 13)

Construct validity (correlate the new measure with some other measure related to it)

Expert judgment (content/facial validity)

Qualitative (Ch. 15) focuses on credibility (member checks, reviews by participants, triangulation of data sources)

Reliability – extent of consistency in results (Ch. 17)

Number of items is important

Objectivity of items is necessary

Qualitative (Ch. 15) focuses on dependability (use more than one person to interpret data, triangulate measures…use interviews plus surveys)

Chapter 8 - Looking Ahead

to Quantitative Data Analysis

Levels of data (nominal, ordinal, interval, ratio) dictate how statistical analysis is done

Nominal (named categories)

Still just names even if numbers assigned as name tags (such as 1=male, 2=female); the numbers have no quantitative meaning

For a single variable, use frequency/percent tables and picture with pie or bar charts (percent=part divided by whole)

For relationship between two variables, use cross-tabulation (also called a contingency table)

For inferences from sample to population, use chi-square test

Ordinal (named categories with relative order, such as class rank: freshman, sophomore, junior, senior)

Analysis is similar to nominal

Chapter 8 (cont.)

Interval (continuous, quantitative measures)

For one variable, use mean and standard deviation and picture with a histogram)

For two variables (both interval) calculate correlation coefficient to see relationship between them; look at direction (positive/direct v. negative/indirect) and strength (weak, r=0 ….. strong, r=1.0)

For two variables (one nominal and one interval), compare means and standard deviations of the two groups

With experiments can compare the net gain/loss from pre to post test for the control v. experimental group

For inferences to pop. use t-tests, ANOVA, significance of correlation

Ratio (continuous, quantitative but with an absolute zero)

Analysis is similar to interval

Chapter 9 - Looking Ahead

to Qualitative Data Analysis

Only interviews are covered here

Qualitative has data collection and data analysis intermingling

Memo writing during data collection (journal of researchers ideas/impressions)

Reflection and reframing of interview questions

Collect until reach data saturation (no new info is being gained with successive interviews)

Chapter 9 (cont.)

Two general approaches

Grounded Theory Approach – inductive reasoning (theories developed based on data collection)

Open coding (track themes)

Axial coding (temporal, causal, associational, valence, spatial)

Core categories developed

Describe process used in analysis

Consensual Qualitative – several people review data and come to consensus about results (auditor checks)

Chapter 9 (cont.)

Analysis techniques

Enumeration (count times theme/construct is mentioned)

Selecting quotations (to support findings)

Inter-coder agreement (reliability check)

Diagramming (see text example)

Peer debriefing (other researchers review/concur)

Auditing (independent review)

Member checks (go back to participants for review)

Identify range of responses

Analyze discrepant cases

Chapter 10 – Preparing a Preliminary Research Proposal

Proposal is a plan for conducting research (the proposal allows feedback on its accuracy) – see TC guidelines

Title – brief statement that names the major variables investigated (and subjects sometimes) as given in the RQ/purpose/hypothesis (see text examples)

Introduction chapter introduces the topic/study by giving the

Problem area/background info (use limited number of references here)

Definitions (conceptual and operational…see text examples)

Significance of study (why it’s important to do the study)

Note that qualitative generally has some info about how the researcher is directly related to the problem (personal experiences and perspectives) since this may influence collection/interpretation of data)

Literature review

RQs/hypotheses

Chapter 10 (cont.)

Methods is the second chapter and as a minimum covers participant selection and instrumentation

Participants – discuss population, how subjects will be selected, number selected, permission, etc.

Instrumentation – describe measurement tools used and cite sources (see text examples and TC guidelines)

Procedures section (explains how the research will be carried out)

Data analysis section (explains how the data will be analyzed statistically)

Threats to validity (limitations/problems with the data)

References/Appendices follow the last chapter

Chapter 11 – Participant Selection in Quantitative Research

Population is the group the researcher is interested in

Accessible population differs from the entire pop, but must be cautious with generalizations from an accessible pop.

Sample is a subgroup drawn from a population

Biased sampling occurs when everyone does not have equal chance of being selected as a participant

Bias can be subtle or obvious

Volunteers, convenience sampling most common biases

Need a system to assure equal chance to everyone – called RANDOM SAMPLING

Even random sampling can have random (chance) errors, but not systematic ones (calculate and report margins of error)

Larger samples have less error

Allows inferences

Chapter 11 (cont.)

Simple random sampling

Names in a hat or using table of random numbers (in book)

Number each pop member with equal digits, then select random starting place in table

Stratified sampling

Divide population into subgroups (and draw from each group…usually proportional)

Systematic sampling…select every nth one from list

Need to make selection through entire list

Beware ordered lists…best to use randomly order lists

Cluster sampling

Population is in groups, ea group is a cluster (Girl Scouts/troops)

Select enough clusters and select them randomly

Chapter 11 (cont.)

Sample size – influenced by several things

Pilot study to help determine return rates

Number of subgroups to be examined (more needed, increases the sample size)

Importance of precise results/need to find a small effect (such as heart attacks…occurrence is small so need larger sample size)

Statistical significance – larger samples have greater chance of significance

See table of suggested sample sizes in book

Chapter 12 – Participant Selection in Qualitative Research

Purposive Sampling (not so interested in unbiased samples so can generalize) – hand pick participants since they possess the needed info

Criterion sampling – very specific criteria must be met to be included; can purchase lists of those meeting criteria or get from professional associations

Random purposive sampling – random selection from a list of those who meet the criterion

Typical case sampling – select what are believed to be normal or typical cases (reduces complexity of data…good for data analysis; beware data don’t apply to most cases)

Chapter 12 (cont.)

Extreme/deviant sampling – identify the criteria (define them) and take from one or both extremes

Useful to refine measurements for later typical case sampling

Intensity sampling – select those with intense feelings or experiences

Maximum variation sampling – making sure participants have full range of characteristics (age/income/gender/etc.)

Homogeneous sampling – opposite of above; want participants to be similar on characteristics

Opportunistic – individuals selected to participate as the opportunity arises since they may have info of interest to the study

Chapter 12 (cont.)

Stratified purposive – identify subgroups then select from them so all subgroups are represented

Snowball (aka chain/network) sampling (also used in quantitative) – for hard to find participants…identify one and ask for referral to others, i.e. drug addicts

Combination purposive sampling – combine any of the above

Improving convenience/accidental sampling – if must be used

Use a larger group selected from multiple locations

Collect demographic data so readers can describe the participants – readers can then judge representativeness

Chapter 12 (cont.)

Sample size in qualitative research

Much smaller than in quantitative (in one published report average N in qualitative was 14 versus 432 in quantitative)

Focus is more on in-depth information gained from participants and requires more time per participant

Saturation sampling – stop when new insights and info are no longer being provided by participants (point of redundancy)

Chapter 13 – Instrumentation in Quantitative Research

General rules for good measurements:

Standardization – use same methods for administering tests to each subject

Objectivity – Use scoring that does not allow subjective interpretation

Social desirability bias – Eliminate this tendency of people to give what they believe is an acceptable answer by using anonymity

Chapter 13 (cont.)

Reliability and Internal Consistency

Reliability refers to consistency in measurement (getting the same reading with multiple measures)

Test/retest reliability – administer test to same group twice (a week or two apart) and should get consistent scores

Measure this using reliability coefficient (0 to 1.0 score and the closer to 1.0 the more consistent); Look for 0.85+

Guessing/random answers; inconsistent administration of tests; changes in subjects; ALL can affect reliability so try to limit these

Internal Consistency from one part of the test to another

Can calculate split-half reliability (odd/even) or Cronbach alpha

On surveys ask positively and negatively worded questions

Chapter 13 (cont.)

Both test/retest and internal consistency provide info that helps in understanding how well a test works

Best to report a couple of reliability measures

Inter-observer Reliability – extent to which 2 observers get the same score, especially important when scoring is not objective

Chapter 13 (cont.)

Validity and its relationship to reliability

Validity is the extent to which an instrument measures what it is intended to measure

Measures can be highly reliable but still not valid

Judgment vailidity

Content – based on expert opinion of the appropriateness of the contents of a test/scale – does it cover the whole content

Facial – on its face does it appear to measure what it’s supposed to

Best to report a couple of reliability measures

Criterion validity – how well measure correlates with some criterion

Concurrent – criterion occurs at same time as measure

Predictive – criterion occurs at a future time (ACT score & college GPA)

Chapter 13 (cont.)

Construct validity – cannot observe some things directly (love, depression, anxiety), thus construct their existence based on observation of behaviors that indicate they exist

Construct validity – refers to the extent to which an instrument yields scores that are consistent with what is known about the construct

Calculate it by correlating scores with other things that indicate the construct ( correl. happiness scale scores to an anxiety scale you developed, should get negative correl.)

Need to have several measures of validity

Chapter 14 – Writing Objective Instruments

Often can find and use existing instruments (ETS Test Collection database)

For this program generally develop your own to develop the related skills…text covers three most common ones

Develop a plan

Have plan reviewed

Revise plan

Write items based on the plan

Have items reviewed

Revise items, put into instrument form

Pilot test instrument

Revise after pilot test

Chapter 14 (cont.)

Attitude scales - Attitudes are orientations toward something and they affect actions

Likert scales are used to measure attitudes by providing descriptive statements and asking whether respondents agree/disagree (SA to SD)

Planning

Identify the components of the object being studied (i.e. job satisfaction made up of working conditions, pay, benefits, etc.)

For each component write 3-5 items plus 1-2 on overall attitude (I like my job

Review/feedback at this stage

Writing

Each statement has only one point (see text example)

Use favorable/unfavorable (neg/pos) wordings

Review items/revise

Pilot testing

Use sample similar to popul. to be tested; think aloud/written feedback

Review/revise

Chapter 14 (cont.)

Observation Checklists – List of behaviors/characteristics for which observations should be made

Planning

Identify the behaviors/characteristics and context for observing them

Have these reviewed, identify new ones/revise

Writing

Each item refers to only one single discrete behavior (no ANDS); see Text

Also often ask about speed, duration, and success of behaviors

Pilot testing

Have at least two people do observations independently and see if they agree on items checked

Review/revise (may need to clearly define behaviors on checklist)

Chapter 14 (cont.)

Achievement tests – Usually objectively scored, multiple choice

Planning

Base test items on objectives – try to cover all or at least most important ones

Have these reviewed/revise

Writing

Multiple choice – one correct choice and at least two plausible distractors

Beware of ambiguous items (see example in text)

Review/revise

Pilot testing

Try to include some high and low achievers

Review carefully and compare performance (of high/low performers)

Revise items

Chapter 14 (cont.)

Note these are only the beginning steps in writing up instruments

It’s a complex process

You can get a doctorate in tests and measurments, so obviously it is a much more complex task than we have just reviewed

You will learn a lot by developing your own instrument for your project

Chapter 15 – Instrumentation in Qualitative Research

Qualitative researchers strive for credibility

Member checks used to verify findings (feedback from participants on correctness of results)

Prolonged engagement in the field gives credibility

Triangulation of data sources (multiple sources give same results)

Triangulation of instrumentation (multiple instruments used for each participant)

Chapter 15 (cont.)

Interviews are most common data collection method, particularly for beginning qualitative researchers

Good source: The Long Interview by Grant McCracken

Pay attention to interviewer selection and behavior so results are not contaminated

Often match interviewers with participants on characteristics, behaviors, etc. (to gain access, trust)

Interviewer self disclosure (interviewer needs to give self-disclosure in both collecting and analyzing data as well as reporting)

Chapter 15 (cont.)

Interview protocols (steps and questions used in the interview)

Simi-structured/open-ended (core questions with some probing allowed)

Initial questions used to establish rapport

Can adopt previously used protocols or develop, review, pilot test your own

Demographic questions should be standardized

Recording responses/note taking – taping is best, but may inhibit responses; make notes immediately, describe process in reporting

Chapter 20 – Experimentation and Threats to External Validity

Experiments: researchers administer treatment to part of a population to determine effects on an outcome variable

Experiments explore cause and effect relationships

Independent variable (IV) in an experiment is cause

Dependent variable (DV) in an experiment is the effect

Simple experiments have one IV and one DV; complex ones have several of either (see text chart)

Chapter 20 (cont.)

Hawthorne effect – the result changes just because subjects are being observed, not because of the experimental treatment

Control groups are used to make sure the HE is not occurring

See text examples before and after use of control groups

Chapter 20 (cont.)

External validity of experiments (generalizability of results to outside the experimental setting)

Threats to external validity

Selection bias – failure to use true random sampling in choosing participants so they are not representative of the population

Reactive effects of exper. setting – a lab setting used in an experiment might differ from the field situation (so different reactions occur in people outside the exp. setting)

Reactive effects of testing/pretest sensitization – subjects in experiment perform differently because sensitive to what effect you are looking for

Obtrusiveness of measurements – try to use unobtrusive measures so no effect on experiment (or do field experiments)

Multiple treatment interference – if several different treatments given to same subjects, can interact

Chapter 21 – Threats to Internal Validity and True Experiments

Internal validity – Did the independent variable really cause a change in the dependent variable (see text example)

True experiments have a treatment and control group…random assignment is essential

Graphically described as:

R O X O

R O O

Chapter 21 (cont.)

Threats include other things besides IV that might have caused a change in the DV

History – any external event that might cause a change in the DV (control for this by having a control group to compare the treated group to)

Maturation – natural changes that occur in subject (control and treatment groups should mature at same rate)

Instrumentation – refers to possible changes in the instrumentation (how the test was administered from pre to post testing); control/trtmt groups should have same problems

Chapter 21 (cont.)

Threats (cont.)

4. Testing (practice effect) – control group should take care of this

Statistical regression/regression toward the mean – those really high (or low) should move down (or up) on subsequent tests; random selection and control group take care of this

Selection – nonequivalent groups compared due to nonrandom assignment (as opposed to random selection)

Can use a Solomon 4-Group design which adds two more groups (one control and one treatmt) that do not have a pretest (to avoid pretest sensitization)

R O X O

R O O

R X O

R O

Chapter 21 (cont.)

True experiments have random assignment – it’s what makes groups equivalent (not the pretest)

Need high external and internal validity to trust results!

Chapter 22 – Pre-Experiments and Quasi-Experiments

Random assignment is the hallmark of true experiments since it controls all threats to internal validity

Sometimes cannot randomly assign so go to the next best

Pre-experimental designs

Single group pretest-posttest design (O X O) - all threats apply

Single group posttest only (X O) - cannot measure change

Static group design – 2 groups, no pretest, no random assignment (dashed line indicates intact groups used)

X O

O

Chapter 22 (cont.)

Quasi-experimental designs

Non equivalent control group design, 2 intact groups

O X O (note that equal pretest scores needed)

O O

Time series design – one group, baseline then treat

O O O O O X O O O O O (observe several, treat, observe)

Equivalent time samples design (X1 trmt and X0 none)

X1O X0O X1O X0O X1O X0O X1O X0O

Quasi- and pre-experimental sometimes used when cannot do true experiments, but need to be cautious with interpretations

Chapter 16 – Descriptive Statistics for Quantitative Qualitative Research

Descriptive statistics summarize data and are used to present data

Quantitative studies – all variables are analyzed statistically

Qualitative studies – demographic variables are analyzed statistically

The level of data determines what kinds of statistical analysis is done (nominal, ordinal, interval, ratio)

The higher levels of data (interval/ratio) allow

more analysis techniques than lower levels

Use measures of center and variation/spread/dispersion

Center: mean, median, mode

Variation: range, IQR, standard deviation

Chapter 16 (cont).

Nominal data – named categories or labels (gender, race, location)

For univariate analysis:

Report frequency/percent (% calculated with part divided by whole); pie charts

Report most frequently given answer (Mode)

For bivariate analysis:

Report frequency/percent by category using

a contingency (aka crosstabs/crossbreak) table

Chapter 16 (cont).

Ordinal data – named categories that have relative order (places in a race, class rank)

For univariate analysis:

Report frequency/percent; bar/column charts

Report most frequently given answer (Mode) and the middle answer (Median) Note: If odd number of scores the median is the middle one, if even number of scores, average the two middle answers

Report the range (highest – lowest) and the innerquartile range (IQR)

For bivariate analysis:

Report frequency/percent by category using a contigency (aka crosstabs/crossbreak) table

Compare medians for different groups

Chapter 16 (cont).

Interval and ratio data – quantitative numbers that tell how much with equidistance between points; ratio also has an absolute zero, so no negative numbers (test scores, age in years, number of minutes late)

For univariate analysis:

Picture distribution with histograms, stemplots

Report mode and median plus the Mean (mathematical average, calculated: ∑x ÷ N) M=population mean; m=sample mean or x-bar

Report the range and SD (average distance of all points from the mean using formula to calculate) SEE TEXT EXAMPLES FOR 3 GROUPS…

larger SD means more variation/spread/dispersion

For bivariate analysis:

Compare distributions

Compare means, SD for different groups

Correlate

Chapter 16 (cont).

Normal distributions – bell shaped and symmetrical

Centerline is the mean (and if normal, also median and mode)

68% of cases within plus or minus 1SD from the mean; 95% of cases within plus or minus 2SD from the mean; 99.7% of cases within plus or minus 3SD from the mean

When have skewed distributions, mean and SD are warped, so use median and range (or IQR) instead

See IQR example in text

Chapter 17 – Correlational Statistics for Quantitative Research

Examine and describe relationships between pairs of scores for a group of participants

Hi/Hi and Lo/Lo is a positive correlation

Lo/Hi and Hi/Lo is a negative correlation

No pattern is no correlation

See tables in text indicating positive v. negative relatioships

Correlation coefficient (Pearson’s ppm r) is a numerical measure for I/R data relationship strength

r is -1.0 to +1.0

sign indicates neg/pos while absolute value of the coefficient indicates strength

Check plot to use Pearson’s r

Chapter 17 – Correlational Statistics for Quantitative Research (cont.)

Text chart on verbiage associated with different values of r

.85 to 1.0 is very strong

.60 to .84 is strong

.40 to .59 is moderately strong

.20 to .39 is weak

0 to .19 is very weak

These refer to absolute values…sign only tells inverse (hi/lo) or direct (hi/hi)

Chapter 18 – Inferential Statistics for Quantitative Research

Inferential statistics used to generalize from a sample to a population

Sampling error may be the reason for relationships/differences found in the sample

Margins of error

Based on number of participants

Larger number of participants, the lower the margin of error

95% confidence intervals is 6.5% - see text example

95% confidence interval is most common; 99% also used

Chapter 18 – Inferential Statistics for Quantitative Research (cont.)

The Null Hypothesis – no real difference/relationship, any difference/relationship is due to random sampling error – not a true difference/relationship in the population

Null hypothesis – no real diff/rel, any found in the sample is due to sampling error

Research/alternate hypothesis – the diff/rel predicted by the researcher

Different statistical tests used based on level of data and which is IV versus DV (see handout)

Chi-square; t-tests; F-tests ANOVA

Chapter 18 – Inferential Statistics for Quantitative Research (cont.)

Generally, probability less than .05 (p ................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download