Analytic Approaches

_edited_edited_edited_edited_edited_edit

Steps to effectively assess outcomes data.

1.

How will your primary outcome be assessed?

Linear regression? Logistic regression? What are your covariates and confounders? How to account for clustered randomization? Random effect model? Fixed effect model? What software will you use?

2.

How will you analyze implementation data?

Click here for a useful toolkit from The World Bank

4.

Check for missing and implausible data.

Examine using a data collector. Is the data quality heterogeneous? Are there ceiling or floor effects? Look for "epidemiological consistencies."

3.

Consider the top 5 variables with large effects in ECD.

These are: age, gender, home situation, socio-economic status, and maternal education.

Watch the video

5.

Data quality checks.

Distribution of scores, inter-rater agreement, test-retest reliability, internal reliability, comparison with norms, correlation with expected variables, differences between testers.