5.2 TIME- SERIES DESIGNS

5.2 TIME- SERIES DESIGNS

time-series design

• series of observations both before & after a treatment

Three types of time-series design

• INTERRUPTED

• NONEQUIVALENT CONTROL GROUP

• SWITCHING REPLICATIONS

INTERRUPTED

• One group is measured many times before and after a treatment (or event).

• Pattern of scores over time is examined.

• Notation: O1 O2 O3 O4 O5 X O6 O7 O8 O9 O10

• Used when a treatment/intervention occurs naturally (e.g., new law, campaign, program).

What does the Interrupted Time Series Design allow us to see

• trends over time → determine treatment effect

• no control group means other factors could also explain the change (→ lower internal validity).

Strongest Evidence of Treatment Effect in Interrupted Time Series Design

• discontinuity (a sudden jump or drop) of score pattern right after treatment

• → clear, abrupt change suggests the intervention caused the change

Why can only abrupt changes be interpreted in Interrupted Time Series Designs?

• because gradual changes are indistinguishable from normal fluctuations over time

What internal validity threats are controlled in a simple Interrupted Time Series Design

• Maturation

• Testing effects

• Regression to the mean

Main threats to internal validity in Interrupted Time Series Design

• History: Another event happening at the same time as the treatment may cause the change.

• Instrumentation: Measurement tools or procedures change over time

NONEQUIVALENT CONTROL GROUP

• Both a treatment group & comparison group are measured multiple times before & after the treatment.

• Notation:
  Treatment: O₁ O₂ O₃ O₄ O₅ X O₆ O₇ O₈ O₉ O₁₀
  Control: O₁ O₂ O₃ O₄ O₅ O₆ O₇ O₈ O₉ O₁₀

• The comparison group does not receive the treatment.

Why is the Interrupted Time Series with Nonequivalent Control Group stronger?

• having a control group measured across the same time period helps rule out history effects (other events happening at the same time).

• This design improves internal validity because both groups experience the same outside influences — but only one group receives X.

• Still no random assignment, so group differences may remain.

SWITCHING REPLICATIONS

• The same treatment is introduced in two different locations/groups, but at different times.

• There is no control group, but the timing difference allows comparison.

• Notation example:
  Location 1: O₁ O₂ O₃ X O₄ O₅ O₆ O₇
  Location 2: O₁ O₂ O₃ O₄ O₅ X O₆ O₇ O₈

Why does Switching Replications improve validity

• if both groups show the same pattern of change only after X, it is unlikely the change was caused by history or chance events.
  → This design strengthens causal inference & supports generalization of the treatment effect.

Quasi-experiments provide

important alternative when true experiments are not possible.

Quasi-experiments lack

• lack degree of control

• lack random assignment

• researchers must seek additional evidence to eliminate internal validity threats

pre-experimental desing

• one-group pretest-posttest design

• bad experiment → little internal validity.

• nonequivalent control group design

• group & a comparison group are compared using pretest & posttest measures

researchers can more confidently make a claim about the effect of treatment when →

• two groups are similar in their pretest scores prior to treatment but differ in their posttest scores following treatment.

controlled threats to internal validity in nonequivalent control group design

• history

• maturation

• testing

• instrumentation

• regression

simple interrupted time-series design

• researchers examine a series of observations both before & after a treatment

Evidence for treatment effects occurs

• are abrupt changes (discontinuities) in the time-series data at the time treatment was implemented.

time series with nonequivalent control group design

• researchers make a series of observations before & after treatment for both a treatment group & a comparable comparison group.