What do you do if your data analysis reveals inconsistencies and errors?

Las inconsistencias en modelo de datos deben ser identificados mediante el seguimiento de kPIS, donde iremos contrastando las desviaciones significativas. O mediante sistemas de backup que nos ayudaran a identificar anomalias en los datos. O generando informes desde las fuentes originales de datos. Por eso es fundamental dise?ar estos sistemas ya que podemos llegar a tomar decisiones erróneas.

Ideally you are not discovering wild inconsistencies and variation in your analysis results. This means that you advanced to the analysis phase of your work without first checking the soundness of the underlying data and the methods you will apply to your analysis. If you are working with a data set that you are unfamiliar with, take time to interview the team that supplied you with the data. Ask them about the data set, methods used for collection, expected ranges of values and descriptive statistics, and any data QA they perform. Once you are convinced that the data are good, then design and conduct analysis. If you eliminate the bad data problem you may design analysis and debug your process if you observe unexpected results.

Implemente a??es preventivas , ao identificar um erro sistêmico é extremamente importante mapear as causas e impactos do desvio . é muito importante a transparência no processo de comunica??o de erro e ainda mais importante o adequado planejamento das a??es de conten??o (imediatas) e de preven??o ( estabelecer barreiras que impe?am a recorrência da falha )

If you believe there are errors in a production data set, you should always seek to interview the team responsible for the data before making changes. The team responsible for the data should be able to explain the condition of the data set, including any known defects or propensity for defects. They should be able to describe normal expected ranges and descriptive statistics for like data sets. If you still find errors you should always retain a copy of your original source data. In fact, if you have access to a version control software it is good to manage data sets under version control. There are many ways you can improve imperfect data sets. If you alter data record your reasoning and methods.

The data should be reviewed to see where inconsistencies may have occurred. If the volume of data is quite large and it is not possible to detect errors visually, it may be useful to get help from additional tools. This help can be obtained by doing the right searches on the Internet or by asking colleagues or experts for advice.

Put in place measures to prevent similar errors in future analyses, such as data validation checks, automated data cleaning processes, or regular audits. Utilize algorithms or scripts to identify and correct missing values, duplicates, outliers, and formatting errors. Regularly run these automated processes as part of your data preparation pipeline before conducting analysis.