Analysing Likert Point Data

Discuss statistics related things
Sofiuu
Posts: 6
Joined: Fri May 06, 2022 11:34 am

Analysing Likert Point Data

Post by Sofiuu »

Hello!
I am working on my dissertation; where I have used a total of 3 measures; 2 of them are 4 and 5 point likert scales, whilst the last is a frequency table: that includes 7 points.- they include Autism Quotient, Food Frequency and Eating Habits. I filtered out data of participants where theyve withdrawn; didnt answer all questions (more than just a handful), for the descriptive table; ive added histograms, box plots, shapiro-wilk, skewness, mode and range (as well as the others that have already been added by jamovi itself), the different responses i got for each questions I had to add to he variables table, and added age, gender and if they were a student in the split by box. My intention is to study the relationship between autistic traits and eating habits, and student (or not) with eating habits.

What sort of analysis should i be doing? And looking at my data; there is a grap for each questions, and my descriptive table has descriptive data for each question also- did i do something wrong here?

I am very new to jamovi and math makes my head all jumbley; if anyone can help; that'd be much appreciated!
User avatar
jonathon
Posts: 2609
Joined: Fri Jan 27, 2017 10:04 am

Re: Analysing Likert Point Data

Post by jonathon »

chi squared test of independence (under contingency table), or non-parametric correlation (spearman or kendall's tau, under correlation matrix).

(also take a look at the surveymv module, its nice for visualising likert items).

jonathon
Sofiuu
Posts: 6
Joined: Fri May 06, 2022 11:34 am

Re: Analysing Likert Point Data

Post by Sofiuu »

Thank you so much!
User avatar
reason180
Posts: 268
Joined: Mon Jul 24, 2017 4:56 pm

Re: Analysing Likert Point Data

Post by reason180 »

I think that if you want to treat your N-point Likert scale data as a close-enough approximation to a continuous, normal distribution, then those variables could be analyzed via standard, parametric methods (t tests, ANOVAs, etc.)
Mishery
Posts: 4
Joined: Fri Jul 05, 2019 5:02 pm

Re: Analysing Likert Point Data

Post by Mishery »

I always point my students to two article to support treating ratings data as continuous:
Carifio J, Perla R., (2008). Resolving the 50-year debate around using and misusing Likert scales. Med Educ., 42(12):1150-2. doi: 10.1111/j.1365-2923.2008.03172.x. PMID: 19120943.
Harpe, S.E., (2015). How to analyze Likert and other rating scale data. Currents in Pharmacy Teaching and Learning, 7(6), Pages 836-850, ISSN 1877-1297, https://doi.org/10.1016/j.cptl.2015.08.001.
Bobafett
Posts: 76
Joined: Thu Jul 18, 2019 11:33 am

Re: Analysing Likert Point Data

Post by Bobafett »

As has been commented on, questionnaire/self-report data from likert scales (i.e. ordinal) is often treated as continuous data when analysing the summed scores, thus availing itself to a wider range of analyses (including, possibly, linear regression that your supervisor may originally had in mind?).

As to your other issue, I think you are asking why you are getting graphs & descriptives per item? You have not mentioned anything about creating/computing a summed total score - and so you are using all the items from the scales as your DVs rather than a single DV per measure - which the summed score would give you.

Apologies if I have misunderstood.
Sofiuu
Posts: 6
Joined: Fri May 06, 2022 11:34 am

Re: Analysing Likert Point Data

Post by Sofiuu »

Bobafett wrote:As has been commented on, questionnaire/self-report data from likert scales (i.e. ordinal) is often treated as continuous data when analysing the summed scores, thus availing itself to a wider range of analyses (including, possibly, linear regression that your supervisor may originally had in mind?).

As to your other issue, I think you are asking why you are getting graphs & descriptives per item? You have not mentioned anything about creating/computing a summed total score - and so you are using all the items from the scales as your DVs rather than a single DV per measure - which the summed score would give you.

Apologies if I have misunderstood.
So you suggest that I change my ordinal data to continuous to get the 'right data'? Would that change any of my other data from the Survey Plots?

Yes! For the second issue; would there be any way to combine my scores so that is wouldn't give me so much?
Sorry if I have somehow made things a little more confusing

And thank you all for the help and advice!
Bobafett
Posts: 76
Joined: Thu Jul 18, 2019 11:33 am

Re: Analysing Likert Point Data

Post by Bobafett »

Changing the type of data from ordinal to continuous in jamovi itself wouldn't do that much really - you'll probably find that you can still run the various types of analyses mentioned above. What I meant was more your reasoning behind this - you might wish to state in your dissertation that you "...treated/viewed the summed data as continuous" rather than any actual change of settings.

There are various ways to create your summed totals in jamovi, but before you do this you need to ensure that none of your items need to be reverse scored otherwise this will mess up the scores. By this i only mean how your questionnaire items are scored e.g. if you are measuring depression on a 1-4 scale then you might expect the higher scores to indicate a higher level of depression such as:
Q1 I feel there is no hope
1 Not at all
2 A little
3 Sometimes
4 Most of the time

This makes sense - you'd expect depressed people to answer 4. But sometimes questionnaires have items that reverse this phrasing in order to make sure people answer consistently and reliably e.g.
Q2 I feel full of the joys of spring
1 Not at all
2 A little
3 Sometimes
4 Most of the time

In this case you'd expect depressed people to answer 1 - but this messes up the scoring and so it needs to be reversed such that 1 = 4, 2 = 3 etc... that way you maintain that those with depression score higher on the items, leading to a larger summed total.

You need to check that the items on your questionnaire(s) do not need reversing. Fortunately there is a very quick way to check - under the 'Factor' icon on the analyses tab you need to run the 'reliability analysis' - dump all your items from a single questionnaire into the 'items to be analysed' window - it'll tell you if there is an issue. If there is no problem, scroll down to the bottom of the test options and under 'Save' tick the 'sum means' box - this creates what you think it will!
If items do need to be reversed, scroll down to the 'Reverse Scale Items' option and put the affected items into the window opposite. Once this is done, then tick the 'sum mean' as before!
Sofiuu
Posts: 6
Joined: Fri May 06, 2022 11:34 am

Re: Analysing Likert Point Data

Post by Sofiuu »

Bobafett wrote:Changing the type of data from ordinal to continuous in jamovi itself wouldn't do that much really - you'll probably find that you can still run the various types of analyses mentioned above. What I meant was more your reasoning behind this - you might wish to state in your dissertation that you "...treated/viewed the summed data as continuous" rather than any actual change of settings.

There are various ways to create your summed totals in jamovi, but before you do this you need to ensure that none of your items need to be reverse scored otherwise this will mess up the scores. By this i only mean how your questionnaire items are scored e.g. if you are measuring depression on a 1-4 scale then you might expect the higher scores to indicate a higher level of depression such as:
Q1 I feel there is no hope
1 Not at all
2 A little
3 Sometimes
4 Most of the time

This makes sense - you'd expect depressed people to answer 4. But sometimes questionnaires have items that reverse this phrasing in order to make sure people answer consistently and reliably e.g.
Q2 I feel full of the joys of spring
1 Not at all
2 A little
3 Sometimes
4 Most of the time

In this case you'd expect depressed people to answer 1 - but this messes up the scoring and so it needs to be reversed such that 1 = 4, 2 = 3 etc... that way you maintain that those with depression score higher on the items, leading to a larger summed total.

You need to check that the items on your questionnaire(s) do not need reversing. Fortunately there is a very quick way to check - under the 'Factor' icon on the analyses tab you need to run the 'reliability analysis' - dump all your items from a single questionnaire into the 'items to be analysed' window - it'll tell you if there is an issue. If there is no problem, scroll down to the bottom of the test options and under 'Save' tick the 'sum means' box - this creates what you think it will!
If items do need to be reversed, scroll down to the 'Reverse Scale Items' option and put the affected items into the window opposite. Once this is done, then tick the 'sum mean' as before!
Thank you so so much!
Sofiuu
Posts: 6
Joined: Fri May 06, 2022 11:34 am

Re: Analysing Likert Point Data

Post by Sofiuu »

Sofiuu wrote:
Bobafett wrote:Changing the type of data from ordinal to continuous in jamovi itself wouldn't do that much really - you'll probably find that you can still run the various types of analyses mentioned above. What I meant was more your reasoning behind this - you might wish to state in your dissertation that you "...treated/viewed the summed data as continuous" rather than any actual change of settings.

There are various ways to create your summed totals in jamovi, but before you do this you need to ensure that none of your items need to be reverse scored otherwise this will mess up the scores. By this i only mean how your questionnaire items are scored e.g. if you are measuring depression on a 1-4 scale then you might expect the higher scores to indicate a higher level of depression such as:
Q1 I feel there is no hope
1 Not at all
2 A little
3 Sometimes
4 Most of the time

This makes sense - you'd expect depressed people to answer 4. But sometimes questionnaires have items that reverse this phrasing in order to make sure people answer consistently and reliably e.g.
Q2 I feel full of the joys of spring
1 Not at all
2 A little
3 Sometimes
4 Most of the time

In this case you'd expect depressed people to answer 1 - but this messes up the scoring and so it needs to be reversed such that 1 = 4, 2 = 3 etc... that way you maintain that those with depression score higher on the items, leading to a larger summed total.

You need to check that the items on your questionnaire(s) do not need reversing. Fortunately there is a very quick way to check - under the 'Factor' icon on the analyses tab you need to run the 'reliability analysis' - dump all your items from a single questionnaire into the 'items to be analysed' window - it'll tell you if there is an issue. If there is no problem, scroll down to the bottom of the test options and under 'Save' tick the 'sum means' box - this creates what you think it will!
If items do need to be reversed, scroll down to the 'Reverse Scale Items' option and put the affected items into the window opposite. Once this is done, then tick the 'sum mean' as before!
Thank you so so much!

Doing what you have advised I can see that there are two boxes: Mean Score and Sum Score. Do I tick the both of them?
Post Reply