Hypothesis Testing
Hypothesis testing – is a process of hypothesis evaluation used in statistics for accepting, rejecting, or modifying an existing assumption. A hypothesis is analyzed by conducting research on a random sample of the population and comparing its results with the initial and alternative assumptions. The alternative hypothesis has the opposite meaning to the initial one, and the research must show which one of them is correct. The choice of methods for hypothesis testing varies according to the type of data and research goals. Also, if the testing demonstrates that the null hypothesis is correct, it also provides it with evidences, which might be of use later.
Process of Hypothesis Testing
Hypothesis testing is carried out on the random sample of the population for avoiding subjective and unreliable results. This sample can be either derived from a larger population or accumulated directly in the process of hypothesis testing.
In order to test an assumption, an analyst has to formulate the null and alternative hypotheses. The first one is the initial hypothesis itself, while the latter has the opposite meaning. In other words, these hypotheses have to be mutually exclusive. If one of them is correct, the other one is automatically incorrect. The alternative assumption is included in the process of hypothesis testing to better display the results.
The process of hypothesis evaluation might be divided into four successive stages:
- Formulation of the null and alternative hypotheses (note that only one of them will be considered as correct by the end of the analysis).
- Creation of the analysis plan.
- Implementation of the plan, or the process of analyzing itself.
- Review of the results (one of the hypotheses will be accepted, while the other one will be rejected).
Example of Hypothesis Testing
Let’s illustrate the term “hypothesis testing” on the following abstract and simplified example. Suppose that an analyst has to understand whether a company’s target audience prefer a product A to product B, or vice versa. He puts forward a hypothesis that the product A is more popular among people than the other product. So, the null hypothesis is that “more than 50% of the company’s target audience will choose the product A over the product B”. The null hypothesis is designated as “Ho”, while the alternative one is “Ha”. The opposite assumption states that “less than 50% of the company’s target audience will choose the product A over the product B”.
When the first stage of hypothesis testing is behind, the analyst can move forward to the next stages – planning and executing the plan. In this example, we skip the description of these steps and jump right to the last one – analyzing the results of hypothesis testing. Imagine that the analyst has conducted a survey about the popularity of the company’s products among its target audience (500 people were involved in the survey). The analysis of the people’s answers has demonstrated that 210 participants chose the product A over the product B, and, correspondingly, 290 of them chose the product B. According to these results, the null hypothesis that initially was put forward by the analyst gets rejected, while the alternative one gets proved.
