I will first use only tests and statistical methods that I am confidant of and that I can defend. Second, if I am that unsure I will post on expert statistics and programming groups like Stack Overflow or write to authors who have used a more robust method. Third, I will ensure credibility of the work overall by making my data and code public on Github.

Esse tema revela uma das grandes barreiras para quem trabalha com dados. No entanto, apesar da escassez é importante o pesquisar lidar com esses desafios e buscar caminhos alternativos que possam gerar resultados satisfatórios, mesmo com essas nuances.

Cuando enfrento limitaciones estadísticas, me aseguro de documentar claramente todas las suposiciones y limitaciones del análisis, lo que refuerza la transparencia. Expongo las restricciones en el tama?o de muestra, calidad de datos o posibles sesgos, para que las partes interesadas comprendan los alcances del estudio. Además, utilizo métodos estadísticos robustos y técnicas de validación cruzada para mejorar la fiabilidad a pesar de las limitaciones. Finalmente, comunico los hallazgos de forma accesible y honesta, destacando cómo estas limitaciones pueden influir en los resultados y qué áreas podrían beneficiarse de datos adicionales o enfoques futuros.

Clearly adopting a Bayesian perspective on the problem. For example if your sample is small you can even estimate the full parametric distribution of a model, with the probability of all the hypotheses, the parameters, quantified.

What I have done when I have presented problems analyzing data because I do not have enough information to generate reliable projections or analysis, is: First, collect the data that I know is reliable and I can use. Second, analyze seasonalities and trends. Third, based on forecasting methods that I know and are reliable. And finally, I analyze the trend and projection of similar cases and pairs, to be able to determine more precisely. It is important to highlight the monitoring and control of the expected vs. the actual data.