If you have several Likert-scored variables, the usual approach is to combine them into a scale and use that as an interval level variable. If you don't want to do that, you could examine the frequencies and see if it is possible to create 3 equal-sized categories. Beyond those options, I cannot think of any method other than your own judgment for converting a set of scores into a smaller number off categories.

David is absolutely right that if there are several Likert-type statements, try to combine them and make a variable/scale. This can be done by conducting factor analysis (Principal Component Analysis or Exploratory Factor Analysis depending on your situation). The value for variable/scale can be mean or sum of all the items loaded on the same factor in factor analysis. Thereafter, as said by David you can treat them as an interval variable. Based on my own experience, I would suggest taking this approach.

Dividing a 5-point scaled statements/items to three groups is very tricky. Some prefer to combined SD and D as low group, Neither/nor point as low group, and A and SA as a high group. However, sometimes this division becomes useless when there are few responses in one or more groups.

If you could tell about your dependent variable, and how many items you have used to measure it, maybe we could guide you better.

Thank you David L Morgan sir and Ali Farooq sir for your suggestions I have updated the question.

With so many items in both your independent and dependent variables, it is unlikely that either of them will form a single scale. Instead, you should begin with Exploratory Factor Analysis to discover the dimensions that you have measured. You will find some useful resources with regard to EFA here:

https://www.researchgate.net/post/How_can_one_help_people_create_simple_scales_from_Likert-scored_items

Thank you Santosh Pal for sharing further information.

I purposely waited for David's response as I knew David was going to refer to the discussion he mentioned in his reply. I was trying to find the link of the same discussion. The link very useful resources which will help you learn EFA, CFA and scale development. I learnt a lot from the same link. Thank you once again David L Morgan :)

The variables in your research have many items, and as David said, it highly unlikely that there will be a single scale for each independent and dependent variable. There must be more than one sub-scales or dimensions for each of two variables. Here are two ways forward:

1. If you have created both of the scales (Social Media Use and Mental Health), the run EFA and divide them into subscale, compute subscale/variables by either taking mean or sum score, and then examine the relationship between subscales of Social Media Use and Mental Health.

2. If you are using the existing scales, that is scales from previous studies, have a look at the studies where these scales were developed. You will be able to find the sub-scales they have created. If sub-scales already exist, a) either run a Confirmatory Factor Analysis (CFA) using the same structure as in the original paper; compute sub-scales as mentioned in No. 1 and run the examine the relationship. b) simply compute Cronbach alpha using the same sub-scale structure as in the original paper, if satisfactory (0.7 or higher), compute variables and examine the relationship. The second approach may not work all the time :)

If you feel overwhelmed with what I have said above, you can also send me a private message and I can guide you through.

Cheers!

Thank you David L Morgan sir and Ali Farooq sir. Your answers have valuable suggestions. I have developed the social media inventory reviewing previous studies and to see the effect of Social Media, I have molded questions in a way that can be related to mental health. In our department people use Mean+SD and Mean-SD to form groups. That is why I was willing to know some new ideas. If you have any further suggestion I will be thankful to you.

Dear Santosh and Colleges,

Really you have a lot of options and a lot of statistical procedures to apply. The principal questions that you have to answer are: what were the factors ore variables that the items of the determinant scales/instrument pretend to reflect? Is this supposed structure validated in your data? After answering this you have to think about what is the best statistical test for dealing with your type of data in a valid and reliable way. Is there not a simple or unique aswer for all this question... whatever you decide, you have to justify in a clear and complete way these decisions make.

Best wishes with your work =)

Dear Pablo Ezequiel Flores Kanter I have developed the social media inventory reviewing previous studies and to see the effect of Social Media, I have molded questions in a way that can be related to mental health. In our department people use Mean+SD and Mean-SD to form groups. That is why I was willing to know some new ideas. If you have any further suggestion I will be thankful to you.

Study topic is Effect of social media on Mental health the objectives are To compare Mental Health of students differing with respect to level of social media usage. To study the effect of gender on Mental Health of students. To study the interaction effect of gender and social media usage Mental Health of students.

The use of Mean + and Mean - SD assumes that your distribution is symmetric, so you should examine the distribution before you use a formula like that. But in general, collapsing a continuous variable into a categorical one "throws away data," so I would not recommend it.