causal research and internal validity

Question 1

CAUSAL RESEARCH AND INTERNAL VALIDITY

Answer 1

descriptive research – gather info (descriptions or measurement) about phenomena
eg. statistics on incidence of eating disorders
causal research – what is cause of illness or other phenomenon
basis to formulate or justify treatment
i.e. provides empirical evidence

Question 2

CAUSALITY

Answer 2

Criteria:
1. antecedence of cause to effect (C occurs before E)
2. co-variation of cause and effect (if C occurs then so does E)
(or if C increases/decreases then E increases/decreases)
3. elimination of rival causal explanations (if C does not occur then E does not occur)

in general establishing lawful functional relationship between cause and effect
however, relationship doesn’t necessarily mean cause and effect
eg. night follows day (predictable, lawful relationship) but day doesn’t cause night
eg. statistical study – correlation between blood cholesterol and heart disease – alone doesn’t prove cause and effect

Scientific research – look to control phenomena – reproduce or change at will
not always possible (eg. night and day)
therefore – describe patterns/co-variations and formulate hypothesis or theory for cause
attempt to eliminate alternative explanations or hypotheses for cause

Question 3

THREATS TO INTERNAL VALIDITY

history
maturation
testing
instrumentation
regression to the mean
selection or assignment errors

Answer 3

compromise conclusion that treatment causes differences or lack of differences observed
i.e. sources of alternative explanation
Threats include:
History – events between pre-test and post-test that are not part of the treatment
eg. initiate low salt diet for hypertension, and patient also starts exercise program
Maturation – time-dependent internal process
eg. infection clearing up by itself
5. Regression to the mean – from unreliability of test measures
eg. choose patient with high cholesterol, second measure more normal
(ie. first measurement abnormally high due to error/fluctuation)
-as increase number of tests, error decreases
6. Selection or assignment errors – groups differ at outset before treatment
eg. assignments not random or low n
7. Mortality (dropout) – if more from one group than other
The dropouts may differ from those that stay, and therefore bias results

Question 4

history

Answer 4

– events between pre-test and post-test that are not part of the treatment
eg. initiate low salt diet for hypertension, and patient also starts exercise program

Question 5

Maturation

Answer 5

time-dependent internal process

eg. infection clearing up by itself

Question 6

testing

Answer 6

practice effects) – repetition of or familiarity with test procedure may change result
eg. familiarity with IQ test may increase performance
Eg. Repetition of test may decreese performance due to fatigue or boredom

Question 7

instrumentation

Answer 7

4. Instrumentation – instrument changes between measurement (reads higher or lower)

Question 8

5. Regression to the mean

Answer 8

– from unreliability of test measures

eg. choose patient with high cholesterol, second measure more normal
(ie. first measurement abnormally high due to error/fluctuation)
- as increase number of tests, error decreases

Question 9

Selection or assignment errors –

Answer 9

groups differ at outset before treatment

eg. assignments not random or low n

Question 10

. Mortality (dropout)

Answer 10

– if more from one group than other

The dropouts may differ from those that stay, and therefore bias results

Question 11

NEED FOR CONTROL

Answer 11

to remove confounding effects of extraneous variables
if can control sufficiently to eliminate other factors has “internal validity”
eg. physicist studying electrical phenomena – insulate against extraneous electrical disturbances
in nutrition research (especially human) sometimes difficult

Question 12

Control group

Answer 12

ensures/increases internal validity
undergo same conditions as treatment group (except treatment)
eg. given inert tablet or saline injection
with people, must also be cautious of placebo effect
control for placebo effect (need non-treatment group)
-no treatment vs placebo vs treatment

at outset need control and experimental groups as similar as possible
“assignment procedure” – method to assign to control and treatment group

Question 13

Control

Answer 13

Control a matter of degree rather than absolute
even tightly designed may allow other explanations arising from unexpected factors
trade-off between internal and external validity
ie. control may generalizability to other situations (ecological validity)
therefore may sacrifice some control (internal validity) for ecological/external validity

Question 14

USE OF CONTROL GROUPS IN CLINICAL RESEARCH

Answer 14

Eg. - redesign, using control group, previous study on exercise in patients with occluded leg arteries
1. sampling – random or matched assignment rather than surgeon’s choice
decrease selection threat
2. pre-treatment test of walking distances for both groups
Decreasing testing threat (since both groups received pretest would be same in both)