evaluation
click to edit
Through the use of a comparison group Control groups: Historical control group (similar to adequacy evaluation design) - internal control group (comparing different intensities of exposure) - External group
Plausibility evaluation
aim to control for confounding factors ie. external factors that can affect the outcome of interest
Has both baseline and endline groups
click to edit
Pros and cons : Not possible to have a true control, Selection bias(women who choose to access the service might be different from those who dont), you cannot rule out alternative explanation for the observed differences, politically and practically easier
click to edit
Probability Evaluation (under RCT,
Does programme P have an effect on outcome Y ( then randomization of P and C will be needed)
Most useful to prove efficacy: What is the effect of biofortified orange fleshed sweet potato on vitamin A
status of children?
click to edit
click to edit
Probability evaluation: Pros and Cons
Probability evaluation: what does it do?
Attempts to ensure that there is only a small and known probability (p< X%) that the difference in Y between P&C is due to confounding, chance or bias
By random assignment of P&C, such that the differences between P&C are minimized
High internal validity
Although high internal validity, external validity (i.e. Generalizability) may below - Stringencies of RCTs may result in situations artificially different from reality
Randomization might be politically or ethically difficult - E.g. large scale conditional cash transfers
Control group contamination or external factors may affect design
A combination of designs is recommended, especially where causal pathways are long and complex
Sample size: what is the change you want to observe, who is being sampled, what is the sampling frame?
Non probability sample: mostly used in process evaluation eg programme manager, Key informants
Source of data, Household interviews
Write a protocol, data collection tools, pilot test, collect data, train data collectors, have data quality checks,
Monitoring indicators, evaluation objectives, evaluation design at least for one component, propose an RCT