Soc. 2155 Week 3 Causation and Experiments I. Causation Relationships between variables Types of association Criteria for causality II. Experiments – testing cause and effect Explanatory research True experimental designs Quasi-experimental designs Internal validity External validity Ethical issues Strengths and weaknesses

Association = relationship Does not have to be causal. Positive association = as X increases, Y increases. Negative association = as X increases, Y decreases. Qualitative variables: presence of X predicts presence or absence of Y.

Which associations could be causal? Years work experience/ income # churches / # bars in a town Cigarette smoking/ lung cancer # firefighters called to fire/ $ amount of damage Race/ poverty Spurious association = apparent association caused by a third factor

Cause = necessary and sufficient condition Necessary: X must be present in order for Y to follow. (ex: to get an “A” it is necessary to complete all assignments). Sufficient: If X occurs, Y must follow. (ex: if you get 100% on every assignment, you will get an “A” in the class.)

3 criteria for causality X causes Y if: X precedes Y in time X and Y are statistically associated All other potential causes of Y have been ruled out.

Additional Criteria Mechanism – connection between “cause” and “effect” – how the cause operates to produce the effect. Context – situations, groups, places, conditions, etc. In which the cause produces the effect.

Determinants/ partial causes Most sociological phenomena have multiple causes. “Determinant” = partial cause or predictor. Not a complete cause. Example: Some determinants of income: Marital status Talent Personality Job duties Type of company Education Skill Training Experience Intelligence Occupation Gender Race Geographic area Industry

Types of Causes Nomothetic Cause – General explanation of a class of phenomena. (e.g., causes of terrorism, crime) Idiographic Cause – Specific event or sequence of events. (e.g., causes of 9/11 attacks, sudden rise in crime rates) May be historical in focus.

Multivariate Relationships X Z X Z Y Y Z intervenes B/T X and Y OR Z “explains” relationship B/T X and Y Z as spurious cause of X and Y Z X X Y Y Z Direct and indirect effects Multiple causes (determinants) of Y

Experiments Explanatory research True experiments Experimental designs Quasi-experimental designs Internal validity External validity Ethical issues Strengths and weaknesses

Explanatory Research Purpose: to explain, to determine cause/effect What is explained? Variation in the dependent variable What can be studied in an experiment? Limited, narrow causal relationships Variables that can be studied in lab Topics for which theory has been developed

True experiment includes Two groups (experimental and control) Random assignment to groups Variation in independent variable (manipulated by researcher) Measurement of dependent variable

The groups Experimental group – is exposed to independent variable (I.V.) Control group - is not exposed to I.V. I.V. is the only difference between the groups Any differences in dependent variable (D.V.) must be due to I.V.

Assignment to groups Randomization Easy to carry out Can control for unmeasured or uncontrolled factors Matching Specific characteristics matched in both groups May be very precise Requires knowledge of relevant characteristics May not control for omitted factors

Pretesting Measures D.V. before experiment Establishes comparability of experimental and control groups Provides baseline for comparison with posttest May teach or “clue in” subjects (pretest effect) Costs extra

Experimental Designs Classic Pretest-Posttest-Control-Group Groups Pretest I.V. Uncontrolled factors Posttest Change Exper. O1 X O3 O3-O1 Control O2 O4 O4-O2 Effect of I.V. = (O3-O1) – (O4-O2)

Experimental Designs Posttest-Only Effect of I.V. = (O1-O2) Groups Pretest I.V. Uncontrolled factors Posttest Change Exper. N/A X O1 Control O2 Effect of I.V. = (O1-O2) Eliminates effect of pretest

Experimental Designs Solomon four-group Groups Pretest I.V. Uncontrolled factors Posttest Change Exper. 1 O1 X O3 O3-O1 Control 1 O2 O4 O4-O2 Exper. 2 O5 Control 2 O6 Effect of I.V. = (O3-O1) – (O4-O2) or (O5-O6) Effect of pretest = (O3-O5) or (O4-O6)

Quasi-Experimental Designs May be used when true experiment isn’t possible Usually involve fewer controls No control group Approximately equivalent control group May take place in the field May be “ex post facto:” designed after the “treatment”

Internal Validity Source of Invalidity Solution History – outside events Control group Maturation – changes in subjects Testing – subject may learn Instrumentation - measurement Statistical regression - moderation Selection bias- groups not comparable Randomization Mortality – dropping out Contamination (competition, demoralization) Treatment misidentification (experimenter expectations, placebo effect, Hawthorne effect) Randomization, double blind, process analysis

External Validity Generalization to “real world” Often a problem in experiments 2 main issues Would sample subjects behave same way outside lab? Cross-population generalizability: would findings hold for different groups, times, places?

Ethical Issues Deception (misleading subjects about purpose of experiment) Selective distribution of benefits (also risks, harm)

Experiments’ Strengths and Weaknesses Isolation of cause/effect High internal validity Easy to replicate Best used for explanatory studies (testing of hypotheses) Weaknesses External validity may be low or undetermined Ethical issues High cost per subject