# Likert analysis of survey data

#### Tahirawan

##### New Member
Hi,

I have performed a survey. The survey is done using a likert scale. The respondents are asked to rate each question on a 5-point scale ranging from 1-5, while “1” represents to “Strongly Disagree”, “5” represents to “Strongly Agree”, and “3” represents to “Neither agree or disagree/Neutral”. This survey is part of a thesis research work for a post graduation study and i need to use SPSS software to analyse the data.

The hypothesis I want to test is to answer below research questions

H1. Does cyber bullying have negative effects on users of social media?

H3. Does social media addiction leads to mental health problems on the user?

H4. How does usage of social media effects the health on the user?

All my research questions are trying to answer a yes / no question

I have around 300 results. So pls. suggest the non parametric test that i need to evaluate the hypothesis? Also if i need to do any goodness of fit test etc to check any assumptions.

/Tahir

Last edited:

#### Berley

##### Member
A couple of thoughts...

You haven't given us enough info to be able to point you in the right direction. What are your groups? Who did you ask?

If you asked both social media users AND non-users, you could use a chi square test to compare the two groups. There are other possibilities as well, depending on how you set up your study.

H4 is not a yes/no question, but you could phrase it so that is similar to yes/no (affects positively or negatively).

#### Tahirawan

##### New Member
The survey was done on university students and includes the following groups

Gender
Age
Citizenship
Ethnicity
EducationLevel
MaritalStatus
EmploymentStatus
MonthlyIncome
DeviceForSocialMedia

Basically all the students were users of Social media so Non-social media users were not part of the study.

I was thinking of doing a 1-sample T test to test the hypothesis questions. For e.g if the mean score of the survey answer is more than 3 then it indicates that we have evidence to say that the hypothesis is true?

For e.g for H2 i have done below analysis.

Survey Questions

1. Do you spent 10 + hours on average on social media sites (Facebook, Twitter , YouTube etc.) on daily basis ?
2. Have you made unsuccessful efforts to control, reduce your Internet use ?
3. Do you think that the Internet addiction has affected your work or study ?
4. Do you feel the need to use Internet more in order to achieve satisfaction ?
5. Do you stay online longer than originally intended ?
6. Do you become defensive or easily offended when anyone asks you about what you do online ?
7. Do you think that life without Internet would be boring, joyless and empty ?

The mean score of all the questions is 3.1 and all answers are significant .05. So we can validate the H2 hypothesis?
Should it be two-sided or one sided test?

See analysis attached for details

#### Attachments

• 76.1 KB Views: 4
• 90.7 KB Views: 4

#### Berley

##### Member
You cannot use a t-test for yes/no questions. If you asked people how many hours they spent online, then you could use a t-test. A t-test is not appropriate or meaningful for yes/no questions.

But the analysis you posted looks more like Likert questions. I'm not sure how you get Likert results from yes/no questions.

If it is Likert data, I would still not run a t-test. I understand some people think it's OK to do so, but I think the general feeling is t-tests are not appropriate for Likert data. But I will defer to others who know more than I.

You could do a chi square (categorical predictors) or logistic regression (continuous predictors) with your data (e.g., is one gender more likely to spend 10+ hours daily on social media; are older respondents less likely to get defensive,..., etc.)