Research Design & Scales of Measurement

Research process:

1) ask research question(s)

2) develop hypotheses

3) collect the data

4) analyze the data

5) evaluate the hypotheses

• Does gun control reduce violent crime? If so, how much?
• What types of people are most likely to develop breast cancer?
• What causes changes in the size of salmon populations?
• How many heroin addicts live in Snohomish County?
• How have academic achievement scores in a school district changed over time?
• sources - personal interest, govt./nonprofit agency, interest group, business, sci. lit.

Developing hypotheses

• specific predictions for empirical research derived from theory or other sources
• one source = theory
• statement of why and how several or more concepts are related
• abstract, applies to multiple empirical cases
• complex or simple
• suggests predictions
• falsifiable
• exploratory research - no specific hypotheses
• accumulated empirical research ---> theory

Collecting data

Populations, samples, and sampling:

• population = set of all individuals, groups, organizations, objects, locations, time periods, or events of interest for your study
• e.g., pregnant women; Fortune 500 companies; adults in U.S.; Superfund sites
• units of analysis
• sample = subset of pop. actually observed
• sampling
• probability - each element or unit of a population has known chance of being selected
• often involves random selection
• nonprobability - not possible to specify the likelihood an element or unit will be selected for the sample
 Population Sample Sample type pregnant women every other pregnant woman at 3 Tacoma clinics Fortune 500 Fortune 500 cos. in WA adults in U.S. random sample of adults in households w/ telephones Superfund sites 65 largest sites in U.S.

• population of interest vs. population to which sample can generalize
Variables and measurement
• variable = characteristic that can take on more than one value
• values = numbers or symbols
• e.g., blood pressure (mmHg), industry (banking/retail/ manufacturing/etc.), Superfund site remediation status (full, partial, none)

Scales/levels of measurement

• determine which statistical techniques are appropriate

nominal - values indicate qualitatively different kinds

• e.g., political party (Demo/Rep/Reform/etc.), soil type (sandy/clay/loamy/etc.)
• qualitative variable = nominal variable
• quantitative var. = ordinal or interval var.

•

ordinal - values indicate ordered categories

• e.g., social class (upper/middle/lower), trimester (1st/2nd/3rd), degree of sunlight (full sun, partial shade, full shade)
interval/ratio - degree of distance/difference between values can be specified
• categorical = nominal & ordinal
• measurement variable = interval/ratio
• discrete vs. continuous

Warning: be careful w/ numeric coding of nominal and ordinal scale data!

Scales of measurement form a hierarchy:

interval

ordinal

nominal

• general rule: use highest scale of measurement possible (most information)

 scale key feature comparisons betw. values interval distance different? which larger? how much larger? ordinal order different? which larger? nominal quality different?

Summary

• research process
• asking research questions - sources
• developing hypotheses
• theory
• exploratory research
• collecting data
• populations, samples, sampling
• variables
• scales of measurement

Part 2:

Data collection methods

• observation – measuring instruments, participant, nonparticipant, unobtrusive
• interviewing - self-administered questionnaires, face-to-face/ telephone interviews, computer-assisted
• surveys
• records/documents/archives

More on measurement

• error - variability in measurements due to unreliability or imprecision
• bias - systematic error - affects all measurements in the same direction
• measured value = actual value + chance error + bias

• univariate - single variable
• bivariate - relationships between two variables
• multivariate - relationships among 3+ variables

hypotheses are predictions about the value of a particular variable or relationships between variables

independent variable = explanatory variable

dependent variable = response/outcome variable

Research design

• experimental vs. observational designs
• experimental - researcher manipulates an independent variable to test causal relationship (does X cause Y?)
• random assignment to one or another group
• practical & ethical limitations to experiments
• observational (nonexperimental) - researcher simply "observes" world
• cross-sectional vs. longitudinal
• cross-sectional = one point in time
• longitudinal = same sample followed over time
• time series

Miscellaneous terms

• cases = elements of/units in sample
• sample size = n = number of cases

Summary

• data collection methods - interviewing, observation, archival records
• experimental vs. observational
• cross-sectional vs. longitudinal