CCJS300

Nominal

• lowest level

• can only be categorized and stated that answers differ from one another

• no rankings. can’t say that one answer is better or more than any other for better or worse

• favorite ice cream, favorite color, yes/no answers

• arbitrary codes (assigning 1 to chocolate for example) the number has no meaning or significance

Ordinal

• second highest level

• answers under this are ranked

• different between value is not precise

interval

• Two characteristics

  • Can compute distances between codes/answers; there is a standard unit of measurement

  • Arbitrary zero point

    • Zero does not equal absence

    • Zero is a baseline

  • Example: Fahrenheit

    • Comparing the difference between 70 and 80 is same as comparing difference between 80 and 90; 0 degrees Fahrenheit doesn’t indicate total absence of heat

Ratio Measurement

• Ratio- highest level, most sophisticated

  • Most versatile, can be used for the most tasks/analyses

  • Two characteristics

    • Can compute distances between codes/answers like interval

    • There is a true zero point

      • 0= absence

      • Example: arrested 0 times, 0 children, 0 dollars

Mode

• The most frequently occurring score

  • No formula is needed to calculate mode

  • Can be used for any level of measurement

Median

• Center, middle score, midpoint

• Splits the distribution into two equal halves

  • 50% of scores are above the median

  • Need to arrange distribution in order (least to greatest or vice versa)

  • Formula- (N+1)2= median position

    • 1 2 3 4 6 7 8 9

    • What is the median?

Mean

• Average

  • Sum X/ N

Reliability

• Consistency and/or stability of a measuring instrument

• Will repeated tests provide the same results

• There are multiple methods to testing reliability

• Test-retest

• Equivalent forms

• Split-half

TEST-RETEST RELIABILITY

• When the same instrument is administered at least twice to the same group, are the results the same (or stable)?

• Give the same test twice to the same subjects at two points in time

• Problems:

• Carryover effect - memory

  • Participants remember what they answered previously

  • The shorter time between administrations

• Answers might actually change in between administrations

  • The longer time between administrations

EQUIVALENT OR MULTIPLE FORMS

• Does the administration of a disguised or alternate version of the same instrument to the same group result in the same (stable) scores?

• Generate two similar, yet different tests given to the same group at two different times • Intended to try to reduce carryover effect
• Both issues, carryover and answers changing, are still present with this

• A third issue: how do we know the items are actually equivalent?

SPLIT-HALF TECHNIQUE

• In the administration of the instrument to one group at one time, are separate halves of the instrument similar in response (are they consistent)?

• Generate two similar, yet different tests that are attached together in one long survey given out at one point in time

• Issues:
  • Maximizes carryover
  • Items may not actually be equivalent
  • Eliminates answers not changing, but this maximizes carryover

Validity

Accuracy of measurement; does the measuring instrument measure what it claims to measure?

Clark and Tifft

• Threatened participants with lie detector tests
  • *Found that after being threatened with the lie detector test, only 18.5% of answers were

changed
• Argued that original answers were pretty reliable

Hirschi

• Used test-retest approach with a group of juveniles and correlated the two tests • *Found that the reliability of self reports depends on the severity of the

offense
• The most serious offenses are the least reliable

Farrington

• Card sort strategy -> gave 14 and 15 year olds cards with delinquent behaviors on it, had to sort the cards into “had done it” or “not done it” piles

• Returns to kids 2 years later (16-17 now) and they had to sort the cards again

• *Found that 25% of the total early confessions were later denied and 50% of the serious early

  confessions were later denied

LAST THING TO REMEMBER ABOUT RELIABILITY AND VALIDITY

An assessment that is highly reliable is not necessarily valid. However, for an assessment to be valid, it must also be reliable.

Simulation

situation or game that attempts to mimic or imitate key

features of a reality

́Stanford Prison Experiment and Milgram’s shock therapy are examples of simulations

́Debriefing- After completion, reassurance of subjects and explanation of the purposes of research provided

Archival or Available Data ́

Form of secondary analysis

́Previously gathered data such as the UCR or ICPSR used for other purposes

́Personal documents or biographies
́Diaries kept by subjects have been used as a source of

information on criminal subcultures

́Respondents are better able to reflect and are able to be flexible in recording events than if studied by another party

́Potential limits and isues
́Data may not contain degree or accuracy or operationalization

wanted by the researcher ́Data may be manipulated

8

wanted by the researcher ́Data may be manipulated

Virtual Reality

́Computer simulation as a means to conduct research or means to

Advantages of Unobtrusive Measures

́Nonreactivity- Research conducted in such a manner that the subject does not react due to being unaware of the researcher’s presence

́Harms that may be caused by the researcher may not occur ́No Hawthorne Effect
́Very cost effective
́Easier to correct mistakes

́Advantageous in examining processes over time or that have occurred in the past

Disadvantages
́

Ethically questionable, could be thought of as an invasion of privacy

́No state granted right to confidentiality or entitlement to privileged communication

́Validity issues may be present depending on the data source ́Scope of study is strictly limited
́Social context is often missing

Mail Surveys:

• Best for: Broad geographic coverage at a reasonable cost.

  Interviews and Surveys.pdf

  
  

• Disadvantages / Issues: Lowest response rate (NN is small); bias toward people with more free time (e.g., retired individuals); no realistic item explanation if a subject is confused; subjects can read the entire survey before answering (affecting ordered items); literacy problems; and subjects may get help from others, reflecting group opinions rather than individual ones.

  Interviews and Surveys.pdf

  
  

• Ideal Schedule: 1st mailing (survey, letter, return envelope) →→ 1 postcard reminder at 1 week →→ 2nd mailing at 3 weeks →→ 7th week final mailing sent via certified mail.

Telephone Surveys:

• Best for: Quick data collection, avoiding literacy problems, and allowing item explanation.

  
  

  
  

• Requirements: Needs to be short (5 minutes max) with short items and few response categories.

  
  

  
  

• Issues / Disadvantages: Limits sample to reachable phone owners; reliance on listed numbers causes bias; screening challenges (hard to reach the correct person); interviewer bias; timing of calls; and asking sensitive items out loud.

  
  

  
  

• Techniques: Random Digit Dialing (RDD) (stratifying by area code/exchange and randomly generating the rest) and Branching (asking a series of questions to narrow down a subject's answer on sensitive items)

Telephone Surveys:

• Best for: Quick data collection, avoiding literacy problems, and allowing item explanation.

  
  

  
  

• Requirements: Needs to be short (5 minutes max) with short items and few response categories.

  
  

  
  

• Issues / Disadvantages: Limits sample to reachable phone owners; reliance on listed numbers causes bias; screening challenges (hard to reach the correct person); interviewer bias; timing of calls; and asking sensitive items out loud.

  
  

  
  

• Techniques: Random Digit Dialing (RDD) (stratifying by area code/exchange and randomly generating the rest) and Branching (asking a series of questions to narrow down a subject's answer on sensitive items).

In-Person Interviews:

• Best for: Achieving the highest response rate, offering immediate item explanation, avoiding literacy problems, and allowing the observation of non-verbal cues (eye contact, body language, appearance of home).

  
  

  
  

• Disadvantages: High interviewer bias (difficult to remain completely neutral), time-consuming, physically demanding, and presents safety issues.

  
  

  
  

• Rules: Maintain a neutral role, be thoroughly familiar with the survey without memorizing it word-for-word, keep a personable demeanor, follow item wording exactly, and record answers precisely

Steps in the Data Management Process

1. Data-Gathering Instrument (e.g., Questionnaire)

2. Codebook (Codes and instructions for coding)

3. Codesheet / Transfer Sheet (Intermediate step)

4. Data Entry (Keyboard entry / Keypunching)

5. Data Verification & Cleaning (Inspecting errors, syntax checks in R)

6. Data File (Storage in computer memory, data disks, or R data frames)

Variable list

Names and definitions for each question or measure.

Editing

Checking errors and inconsistencies.

Coding

Assigning consistent numeric or category values.

Data verification

Using software (like R) to inspect and validate data.

Historical methods

Keypunch cards and optical scanning.

Modern methods

Importing CSV (read.csv()), tab-delimited (read.delim()), or fixed-width DAT files (read.fwf()) directly into R.

Types of Data Collection

• Surveys, interviews, self-reports, observation, and secondary/archival analysis.

Exploratory / New phenomena

Used when examining a new or unfamiliar topic or problem.

Descriptive / Old phenomena

Used to document characteristics, trends, or established populations.

Longitudinal

examines processes or changes over time (e.g., time series designs).

Cross-Sectional

Cross-sectional looks at a population at a single point in time;

Classical Experimental Design (CED)

Requires 3 essential elements:

Experimental and Control Groups

Experimental gets the treatment; control does not.

Equivalence

Groups must be equivalent. Maintained via Random Assignment (handles systematic error via chance; random error is accepted) or Matching (cloning/twinning individuals based on key characteristics like age, gender, race, or crime committed).

Pre-tests (O1​) and Post-tests (O2): Measuring before and after treatment (T).

Pre-tests (O1​) and Post-tests (O2): Measuring before and after treatment (T).

Solomon Four-Group Design

Combines CED and posttest-only designs across four groups to check for pre-testing interaction effects.

Quasi-Experimental Designs:

One-Group Ex Post Facto (T,O2​)

One-Group Before-After (O1,T,O2​)

Two-Group Ex Post Facto (T,O2​ ,vs. O2​)

Time Series Design

Uses multiple pre-tests and post-tests to track natural fluctuations and identify true policy/treatment impacts (e.g., tracking crime monthly before and after adding police presence).

Counter-Balanced Design

: Created to handle multiple treatment interference by exposing groups to every possible combination or order of treatments (e.g., A, B, AB, BA, and No Treatment groups).

Double-Blind Experiment

Neither researcher nor subjects know who gets the treatment; a placebo and a third-party tracker maintain blindness.

Operation 25

: Single-group study testing police presence on crime rates (lacked equivalence).

Rand 20th Precinct Study

Two demographically similar NY precincts, increasing crime enforcement in only one.

Kansas City Preventive Patrol Study

15 beats split into proactive, reactive, and control groups to test patrols on crime and fear of crime

Berk and Sherman Domestic Violence Study

Randomly assigned arrest, forced separation, or mediation to domestic violence calls to measure deterrence.

Simulations

Situations or games that attempt to mimic or imitate key features of a reality (e.g., Stanford Prison Experiment and Milgram’s shock therapy). Requires a Debriefing afterward to reassure subjects and explain the research purpose

Scenarios, Vignettes, and Parables

Used in research to present hypothetical situations to respondents to gauge attitudes,ethical choices, or behavioral projections.

Simple Observation

Researcher makes no interference; watches people in natural environments.

Naturalistic observation

: Directly in the occurring environment; no interaction.

Structured observation

In a lab or simulated environment to observe a specific, limited set of behaviors.

Disguised Observation

Informed consent is not sought; researchers pretend to be part of the study group with complete immersion.

Confederates

Individuals planted within a study to facilitate or provoke desired behaviors.

Advantages of Unobtrusive Measures:

Nonreactive (no Hawthorne effect), no researcher-induced harm, highly cost-effective, easy to correct mistakes, and great for examining historical processes over time.

Disadvantages of Unobtrusive Measures:

Ethically questionable (invasion of privacy), no state-granted confidentiality/privileged communication, potential validity limits depending on data source, limited scope, and missing social context.

Archival / Available Data

Form of secondary analysis using previously gathered data (e.g., UCR, ICPSR) or personal documents/diaries.

Limits: Data may lack desired degree of accuracy or operationalization, and data may be manipulated.

Physical Trace Analysis

Indirect study of deposits, accretion of matter, and physical evidence left behind by previous human interaction.

Content Analysis

Systematic classification and study of mass media (newspapers, magazines, television).

Validity Definition

The accuracy of measurement; does the instrument measure what it claims to measure?

Face / Logical Validity

Subjective, simplest judgment of whether an item appears to measure the concept.

Content Validity

Ensuring individual items comprehensively measure the concept across multiple indicators (non-empirical, though item analysis can eliminate non-discriminatory items).

Construct Validity

Accuracy of scale in tapping the correct theoretical concept; uses empirical methods like factor analysis or comparing against prior research.

Concurrent Validity

Assessment correlates with other existing assessments measuring the same construct (A correlates with B).

Predictive Validity

Assessment accurately predicts performance on a future criterion (A predicts B).

Pragmatic Validity

Accuracy in predicting current or future status.

Convergent-Discriminant Validity

: Different measures of the same construct should yield similar results (convergence); the same measure of different concepts should yield different results (discrimination).

Triangulation

: Using multiple methods of data collection (e.g., Barbara Heyl's study of a brothel madam using observation, interviews, content analysis, and life history data).

Major Surveys

National Crime Victimization Survey (NCVS).

Limitations

Relies on human memory, subject to telescoping, excludes victimless crimes or murders, underreporting due to fear or embarrassment, and potential misclassification of incidents.

UCR

: Uniform Crime Reporting Program (FBI-compiled official crime data).

NCVS

: National Crime Victimization Survey (Bureau of Justice Statistics household survey).

ICPSR

: Inter-university Consortium for Political and Social Research (archive of social science data).NSDUH: National Survey on Drug Use and Health (tracks substance use across 50 states).

NSDUH

National Survey on Drug Use and Health (tracks substance use across 50 states).

Self-Reports & Associated Issues

• Origins: Originated with James Short and Ivan Nye (1957–1958) studying juvenile delinquency and parental SES.

  
  

  
  

• Types of Self-Reports: Opinionaires, checklists, questionnaires, interest inventories, attitudinal scales, and interviews.

  
  

  
  

• Devices: Woodworth personality test, Bell's Adjustment Inventory, Minnesota Multiphasic Personality Inventory (MMPI), Ahluwalia's Teaching Attitude Inventory.

self report Advantages & Disadvantages

: Quick, cheap, high honesty on direct questions, easy to code; but limited to juveniles/males historically, fixed questions lack flexibility, and prone to severe honesty/accuracy issues.

Specific Biases & Recall Issues:

• Lying: People lie by under-reporting or over-reporting offenses.

  
  

  
  

• Telescoping / Backwards Telescoping: Difficulty recalling time frames; when asked about events in the past year, respondents shift older events forward into the timeframe, making rates and offenses appear artificially higher.

Reliability

Consistency and stability of a measuring instrument; will repeated tests provide the same results?

Test-Retest

Giving the same test twice to the same group at two points in time. Issues: Carryover effect (memory) or actual behavioral changes between tests.

Equivalent / Alternate Forms

Giving two similar yet different versions of a test to reduce memory carryover. Issues:Carryover still possible, answers can change, and establishing true item equivalence is difficult.

Split-Half Technique

: Administering one survey at one time, splitting items into two halves, and checking consistency. Issues: Maximizes carryover since items are in one survey.

The Golden Rule

: An assessment that is highly reliable is not necessarily valid. However, for an assessment to be valid, it must also be reliable.

read.fwf()

reads fixed-width format files (like Delinquency.dat) using designated character widths.

table(data$SEX, useNA = "ifany")

checks category counts and missing values.

factor() and ordered()

handle nominal and ordinal data properties in R data frames.

Importance of r synax

Keeping raw data unchanged and documenting every transformation in a reproducible R syntax script ensures data integrity and error correction.