Tap to Read ➤

What is a Two-tailed Test? We Bet You Didn't Know This

Ishani Chatterjee Shukla Apr 29, 2019
A two-tailed test is not a mathematical freak (similar to freaks of nature such as two-headed snakes!), and neither is it that hard to tackle! Let's find out all about this statistical tool, through this post.

The Two-tailed Test

Before letting out the details of the two-tailed test, let's take a quick look at statistical inference. A statistical inference refers to the process of drawing conclusions from a given set of data that may have an influence of random variations, such as sampling variations and observational errors.
A two-tailed test is a statistical test which is used to arrive at a statistical inference. In this test, the null hypothesis H0 is rejected in case the value of the test statistic, which is a function of the sample, is either sufficiently small or sufficiently large.
This is where the one-tailed test vs the two-tailed test difference comes into picture, as in case of the former, only one among the two rejection areas (either sufficiently small or large) is preselected in accordance with the fact that the alternative hypothesis is chosen, and rejection takes place only in case of the test fulfilling this criterion.
When you draw a bell curve, which is what a normal data distribution curve is known as, you get two data tails at the horizontal extremities, parallel to the X-axis. It is due to these data tails that this test gets its name.

When to Use This Test?

This test is used for statistical inference when the statistical hypothesis is rejected due to its value being either sufficiently large or sufficiently small. In other words, when the null hypothesis gets rejected due to its value falling under either tails of the sampling distribution curve, a two-tailed test is used to detect changes in the parameter.

The Formula

The formula for drawing statistical inference using this tailed test goes as follows:

ɀ = (Arithmetic Mean-μ)/(σ/√n)

ɀ = test statistic
μ = mean of population values
σ = standard deviation of differences between all pairs
n = total number of pairs
Therefore, using this formula, we can take a look at the following example to better understand this statistical concept.
During a recruitment drive, 150 job applicants were asked take a test to evaluate their Intelligence Quotient (IQ). From the data provided by the applicants, the recruitment personnel found out that the mean IQ is 120, and the standard deviation of the results is 11.3. They also estimated that the population mean would be less than 116.
Use a 0.05 significance level and decide whether the recruiters are correct in their hypothesis.

H0 = 116
H1 = >116
Therefore, using significance level 0.05, critical value = 1.645
Since, ɀ = (Arithmetic Mean-μ)/(σ/√n),
Hence, ɀ = (120-116)/(11.3/√150) = 4.34
We can see, on plotting the bell curve, that the test statistic 4.34 will fall to the right of the critical value 1.645, falling inside the critical region. This rejects the null hypothesis H0.
Therefore, it can be concluded from the two-tailed test that the alternative hypothesis H1 has been accepted, citing the value of the population mean to be greater than 116. Hence, the statistical verdict contradicts the recruiter's estimate.