In this discussion, you will cover what it means to be statistically significant, or, in statistical terms, . When you are analyzing data and you want to know if your findings are significant you will use a P-value. Unfortunately, statistical significance is often misunderstood in organizations. And because more organizations are relying on data to make critical business decisions, its an essential concept for you to understand.
An analysis is statistically significant when the relationship between the variables cannot be explained by chance alone. For example, imagine that you are trying to determine the cause of an organization’s turnover rate. When you encounter this problem of employees leaving and hypothesize answers numerically, you can then subject the answers to hypothesis testing. This testing can determine the probability of observing these results under the null hypothesis by random chance. In the absence of numerical data, you may attribute all turnover to low unemployment rates. However, by adding numerical answers, you can determine the significance, or likelihood, of unemployment rates being the cause. Therefore, a value less than .05 is statistically significant. If the P-value between the turnover and unemployment rates is less than .05, then you can be more confident that the two are correlated. The closer the p-value is to zero on the y-axis, the more correlated the two variables are. See the image below for a visual representation of this.
, therefore, are used as to alleviate uncertainty from the truth, if interpreted accurately. However, there is a widespread reproducibility crisis occurring in the scientific community, which unearthed inappropriate application of statistical methods across a massive amount of research publications. This inappropriate application causes irreproducibility of results. P-values are a pervasive problem because they are misapplied, they answer questions that are not asked, and they are quite misunderstood. To make matters worse, p-hacking, or manipulating data to derive statistically significant results, has become common. When you are using P-values to review the correlation between variables, it is important to ensure results can be replicated in the future.
Discussion Topic
As you learned above, it is critical to understand what the business analysis results indicate. In this discussion, you are going to dissect how to spot a statistical storyteller.
In your original post, answer the following:
- How can you detect if a study actually has statistically significant results?
- Why would a business analyst manipulate results to derive a p-value of <.05?
- Research and summarize a study that used data manipulation or p-hacking. What was the red flag that helped readers spot the statistical storyteller? (NOTE: If you need help, check out these .)
