CHI-SQUARE TEST; LET’S EXPLORE WHICH FACTOR IMPACTS THE DECISION FOR ENTERING INTO TRIAL LESSONS AT THE COMPANY THAT I AM WORKING FOR?
Hi everyone! The company that I am working for, helps its students to practice in English via Skype with native speaker teachers. Our programs might be in 3,6 or 9 months periods. The teachers at the company call their students for 10, 20 or 30, and more minutes daily on weekdays. Before people buy our language programs, they have a chance to attend a trial lesson. In this article, I will show you that how I tested which factors stored in our database are related to candidates’ making decisions to join our trial lessons.
After the trial lessons, our education advisors tell about our programs to the candidate students after they apply to get information about the programs. Also, the candidates can apply to a trial automatically while they are filling the online form for buying our programs. The system is designed to get a trial lesson on the company’s mobile application. Some of the candidates are eager to enter this trial session, but the others not. The form applicants who had the trial lessons are likely to buy our service.
We launch an automatic system for trial lesson appointments then, we try to optimize the ratio of attendance in a shot of our language learning method; consequently, we try to increase the sales of our company. To optimize this trial session’s ratios, we checked whether some dimensions are related to test call or not. Let’s assume these dimensions like that how many foreign languages candidates can speak? Secondly, have they ever had an online course? Lastly how often do they travel abroad?
All these data are categorical. Also, the attendance to test sessions is a kind of categorical data too. To examine this categorical data correlation, I used the chi-square test.
A chi-square test can be used as a non-parametric test. It means we do not check the distribution of data set before we apply the chi-square test. There are three ways of practicing the chi-square test. These are the goodness of fit, the test of independence and test of homogeneity.
In this business problem, I want to know is getting a trial lesson up to the three dimensions explained above. We can use any statistical tool that has a setting for the chi-square test.
Let’s start with checking the effect of the number of students’ foreign languages on attendance in trial lessons. We asked people how many languages they can speak. Their answers are categorized as “I cannot speak any other language.”, “I can speak only English as a foreign language.” and “I can speak more than one foreign language.” My null hypothesis is the attendance in the shot of the company does not depend on speaking a foreign language.
We can see the cross-tabulation for the chi-square test in Figure-1. This is a frequency table for people’s responses and their status of the trial lesson. We can see the expected values are half of the numbers in each category. There is no big difference between categories in terms of trials.
The table shows the chi-square values in Figure-2. I get p=0,05 for the significance level. I see the significance of chi-square is higher than 0,05. Therefore I will accept the null hypothesis. These two components are independent. The candidates who can speak other foreign languages do not enter the trial sessions more or less.
Secondly, I will test the familiarity with online courses and the eagerness for participating in our trial sessions on the phone. The hypothesis test again emphasizes independencies.
On the cross-tabulation figure, the familiarity for online courses has two major groups. The first group is in favor of trial sessions and at the same time, they are familiar with online courses. The second group has an opposite profile in terms of trials and online courses. They did not attend both.
According to the Figure-4, these two variables are dependent. The Pearson Chi-Square significance is less than 0,05. I accepted the alternative hypothesis. People who already had online courses before are eager to enter trial sessions.
Lastly, my null hypothesis is the attendance in the trials and the frequencies of traveling abroad are independent. The alternative hypothesis is these two variables are dependent. Let’s check which thesis that we will accept after the chi-square test.
Here in Figure-5, the candidates responded to the question about traveling abroad in four categories that are “rare”, ”sometimes”, ”often” and “very often”. There are so many candidates in the “rare” category who did not attend the trial. The category of “sometimes” has also many candidates who did not prefer to take trial sessions.
The significance level is 0,05. Pearson Chi-Square significance is 0,01. This is less than the p-value, so we can reject the null hypothesis and accept the alternative hypothesis. Traveling abroad has an impact on entering our free trial sessions.
In conclusion, now we know which factors have an impact on the attendance in trial sessions. We will do some upgrades and modifications in our automatic trial system.
Resources:
Chi-square test in business research, Dr. Rakesh Kumar, Shilpi Sharma, May 2016
