Missing data

Em minha pesquisa de tese, graças, a priori, à falta de eficiência dos tribunais de contas, tenho muitos dados faltantes (missing data) em meu banco de dados, gerado a partir de indicadores fiscais, orçamentários e de transparência de 282 municípios brasileiros (todos com população superior a 100 mil habitantes).

Para clarear as ideias e lidar melhor com esse problema, estou lendo o livro Missing data: a gentle introduction, dos autores Patrick E. McKnight; Katherine M. McKnight; Souraya Sidani; Aurelio José Figueredo, publicado pela editora Guilford Press, 2007 (New York).

Encontrei algumas citações interessantes no livro, que gostaria de partilhar no blog. Inicialmente pensei em fazer uma tradução livre, mas por falta de tempo e para oferecer aos leitores o texto original, segue!

"As the old saying goes, the only certainties are death and taxes. We'd like to add one more to that list: missing data. As any social scientist can attest, missing data are virtually guaranteed in research studies."

“The best solution to handle missing data is to have none.” (R. A. Fisher)

No Prefácio do livro, os autores agradecem a um colega chamado Lee Sechrest, que colaborou com sugestões valiosas para o livro. Ainda de acordo com os autores, Lee propôs quatro “leis” que foram recentemente publicados por Bradley Smith (2006) e essas influenciaram sua forma de pensar, de forma que decidiram apresentá-las no livro. Eis!

Lee’s Law Number 1: Everything eventually turns up.

This law says that missing data often are not missing. If the data are simply misplaced on our desk, a thorough search may very well turn up data we believed to be missing.

Lee’s Law Number 2: More is better than less.

We view this law not from the perspective of missing data but from the perspective of data. If we have more actual data, that can offset missing data. More missing data, however, we would not want!

Lee’s Law Number 3: Rarely is there such a thing as a true catastrophe.

Unless all information is lost, most missing data situations can either be prevented or treated. Catastrophic data loss is rare enough that we do not need to prepare ourselves for that outcome.

Lee’s Law Number 4: Nothing in statistics or research methodology requires you to do something stupid.

This last law is our guiding principle. Thoughtful approaches should always be preferred. No procedure warrants using unsound data-handling methods. People who have misused or performed contraindicated analyses should own up to the mistakes and not blame an approach. At the heart of it all, we embrace a philosophy where researchers calmly approach missing data as an interesting question rather than as a catastrophe. The researcher acts as a detective, “solving” the problem without compromising the integrity of the study.