RESEARCH DESIGNS

IV = Intervention occurs (the level of the independent variable is changed)
DV = Dependent variable
tn = Time of measurement of DV
X = Experimental group
C = Control group

Desirable features in research designs:

1.  Premeasure the dependent variable
2.  Experimenter controls or applies the intervention (independent variable)
3.  Use control group, with random assignment (best) or matching

EXPERIMENTAL DESIGNS

Before-and-after (classic experiment)

Every desirable feature

(X) DVt1…......IV.......DVt2
(C) DVt1…….........…DVt2

Solomon four-group design

Even better - checks for effects of premeasure
 
(X1) DVt1 …. IV.…..DV t2
(C1) DVt1 ………… DV t2
(X2)....……... IV...... DV t2
(C2)....……………...DV t2

Effect of the intervention:  (X[DVt1-XDVt2]) – (C[DVt1-CDVt2])

Experimental panel design (interviews) or "interrupted time-series" design

Longitudinal equivalent of the classic experiment.  Dependent variable measures can reflect interview responses or other data (e.g., crime rate)

(X) DVt1…IV…......DVt2…......DVt3…......DVt4
(C) DVt1………......DVt2…......DVt3…......DVt4

QUASI-EXPERIMENTAL DESIGNS

Before-and-after, but lack a control group, or use a control group that is not matched or randomized

DVt1….....IV…......DVt2

Same, but design is a "panel" (dependent variable reflects interviews) or a "time-series" (dependent variable measured several times)

DVt1…....IV…......DVt2….....DVt3….....DVt4

After-only, with or without a control group

No measure of the DV was taken before the independent variable was applied

With control group

(X) IV…DV t2
(C) IV…DV t2

Without control group

IV…DV t2

Non-experimental design that mimics an experiment

  • A known "intervention" - the independent variable was adjusted on a certain date
  • There are measures of the dependent variable before and after the date
  • Preferable - a control group that closely matches the characteristics of the experimental group can be identified

NON-EXPERIMENTAL DESIGNS

Investigator does not adjust the independent variable.

Ex-post facto/"passive statistical"

Most common type of CJ research, where both the dependent variable (DV) and independent variable (IV) are measured retrospectively (after the fact).

Cross-sectional

Subgroups of the independent variable (e.g., gender) are separately assessed for their effect on the dependent variable.

Panel or time design

Repeated interviews of the same group of subjects over time, or repeated measurement of a single characteristic over time (e.g., crime rate).  These are often called "cohort" studies.

T1…T2…T3…T4

Multiple-group trend

Larger group is broken down into matched or randomly drawn subgroups, and each is interviewed once over time.  The goal is to control for the influence of the measurement process.

subgroup1 ----T1
subgroup 2 --------T2
subgroup 3 ------------T3
subgroup 4 ----------------T4

 

Home