Please enable JavaScript.
Coggle requires JavaScript to display documents.
L8 - Biostatistics contnuedProblems in Research To describe and recognise…
L8 - Biostatistics contnued
Problems in Research
- To describe and recognise common problems in research including: intervention effect; restricted ranges; violating the independence of observations; mistaking correlation for causation; unequal groups; intra-group dependency, and external validity issues.*
Effect of Intervention
Failing to discriminate between intervention effects at the group level and at the single-subject level is a common mistake.
A statistically significant effect at the group level does not mean you should assume every person has benefited.
We use a Reliable Change (RC) Index
This indicates the degree to which that partcular person has been changed by the outcome measure during the course of the intervention
If RC > 1.96, it is likely there is a real change between pre-test score and post-test score (indicating intervention effect)
likely that the post-test score is reflecting
a real change rather than measurement error
-
Mistaking Correlation of Causation
To reach a conclusion that one variable causes another, three criteria must be met:
-
-
Causal closure -
All other variables that might influence the relationship between the two variables must be eliminated or controlled.
Requires Equivalent Groups
Violating the Independence of Observation
- The independence (of observations) assumption states that each score on the dependent variable must be independent of another score.
- Violations of the independence assumption will inflate the Type I error rate
Unequal Groups
Two groups are equivalent if they are matched on all extraneous variables that might otherwise produce between-group differences on the dependent variable. If unequal, cannot attribute differences solely to the independently variable
Selection Bias triggers unequal groups
-
-
Intra-Group Dependecy
It must be expected that there will be similarities within groups, based
on environmental conditions.
Within groups, scores will be statistically dependent
External Group Validity Issues
Research findings have external validity if they can be generalized to the wider population.
If we randomly select our sample from the target population, then our sample will be representative of the individuals in the target population
