Data analysis is one of the most useful tools that our humanity has created. Data analysis gives us an enormous amount of valuable data in a short amount of period. It enables us to predict the future and helps us make a better decision. However, when we encounter these analyses, it is always important to understand the whole picture and to avoid creating biased conclusions. As the internet has become part of our daily life, we encounter more data. However, because of the fake information that can be created with stakeholder analysis, it is nowadays very important for us to understand common misconceptions and biases that can be created by the people who are doing the analysis.

To begin, have you seen research from a pharmaceutical company that advertises that people who constantly consume their health products have an average of 10% more lifespan compared to average people? While these researches are most likely be true, the factor that we have to consider is called a confounding variable. Confounding variable is a factor that effect both variable.For the example above, people who consume health products daily are more motivated to maintain better health. This means that while the health product from the company could have possibly had a positive effect on the person, this result is probably from other factors too, such as a healthy sleep schedule, healthy food, and exercise. Without thinking about the confounding variables, we can even conclude something like that ice cream sale increases shark attacks which we can think from our common sense that they don’t have any correlation or a mutual relationship. The confounding variable that is creating this result is the season, where in summer, many people eat more ice cream, and it is also true that in summer people frequently visit a beach, which causes more shark attacks. It is very important to understand if there is an actual direct relationship between two factors, and it is important to check if there is any other confounding variable that is causing biased information.

Next, the survey is an essential and useful tool for understanding the ideas of a large group of people. While a survey is very useful, a survey is one of the easiest ways to create a fake analysis. The most important thing about a survey is whether it shows the whole population. Collecting surveys from part of the population that has certain traits can cause biased information. For example, if a company surveys employees who are in high positions and highly paid. They’ll have more chance of feeling happy about their company. It is important to understand whether surveys represent the whole population. 

Third, large numbers are sometimes confusing, so percentages are very useful most of the time. However, because percentages are ratio, they often don’t represent the size of the data, which can cause misconception and biased conclusions. For example, if we think of a situation where the acceptance rate for Companies A, B, and C equals 60%,30%, and 80% for women respectively, and 40%,70%, and 20% for men respectively. Many might assume that women had the advantage of getting into the companies because women had a higher chance for two companies, however, we can never conclude this. This is because the percentage doesn’t show the size that the percentage is representing. 10 candidates out of 5 and 1000 candidates out of 500 are considered the same percentage. So for the example above, only about 100 candidates could have applied for companies A and C, and more than 1000 could have applied for company B, and this would create a higher chance for men overall. This error that most make is also famously known as Simpson’s paradox, which states that the relationship between two variables could be different from a partial relationship when one or more variables are controlled. For the case above, the numbers are not faked but it created a piece of biased information. While it is true that percentages are easy to process, percentages sometimes can be misleading. It is very important to check the size that percentages are representing before comparing multiple percentages. 

Fourth, have you seen those ads that their product’s effectiveness increased 100% from last year? This is another type of way that companies use to trick their customers. The main point is that it has been increased 100% from last year. In the sentence, it doesn’t contain any information about last year. This means that if the value of last year was very low, just even a small increase can be shown as a massive increase. For example, if only 0.5% of the population in City A supported Law A last year, and this year it increased to 2% of the population in City A supporting Law A, while this is now only from 0.5% of the population to 2% of the population, the group who support Law A can still advertisement as 400% growth. The main key to looking at data with percentages is to understand what the percentage truly means as a whole. 

Last, but not least, while data analysis is a very useful method that shows data without explaining much about it, however, these analyses sometimes don’t show the true picture of the data and can be used to create false information. As more data is shared and read every day, it is our job to understand some of the most common traps of data analysis and to make a clever decision. I hope this magazine helped you better understand how to interpret data.