Enormously valuable explanation! Thanks Anthony! And I'm glad to see that you have made lots of other (undoubtedly) presentations! I look forward to learn from them.
Hi . thanks for that video. have a question: on 11:37. what are the statistical test or analysis we may run to find the correlation between these 4 domains of health satisfaction? suppose 3 independants variables and 1 dependant variable as they have been measured by likert scale
Hi, thanks for your question. The WHOQOL-BREF produces four Likert scales (the 4 domains of QOL). If you wanted to know the correlation or covariance between these four domains, you would likely use Pearson's correlation coefficient (r). However, if the sample size is small, you would not meet the required assumptions for Pearson's, so you could use non-parametric correlation tests, such as Spearman's rho or Kendall's tau.
Hi @Anthony, thank you so much for this clear explanation. Could you please indicate me what video of your channel follows this one? At minute 15:06 you mention that session 19 will deep on this topic. thank you in advance
Hi there and thank you for your interest! On the channel if you go to playlists you should be able to find Session 19: Inferential Statistics Which has 8 videos on topics like Z score, normal distribution, t tests, etc.
Hi Anthony, Thanks very much for this helpful lecture. I am a student and do not have much experience with data-analysis and SPSS. I hope you can help me with the following question. I am using a Likert subscale with 3 questions with values 1: strongly disagree up until 5: strongly agree. What I would like to do in SPSS is to compute the mean values of these questions and label them as ordinal variables. My teacher said that that is an option if you do not have a lot of values. Possible values would be then: 1, 1.5, 2, 2.5, 3, 3.5, 4, 4.5 and 5. What would be logical names to use for these computed values? Thanks in advance, Pieternel
Hi Pieter, it's an interesting question. I've never encountered that approach, so it's hard for me to comment. In terms of creating logical labels, I suspect you would need to do some research to see what people would associate with a term that is "4.5." In other words, what do people generally consider a level of agreement that is part way between "agree" and "strongly agree." Thinking of how I would answer that question myself, I cannot come to an answer. I'd be curious what you decided. I'm assuming the purpose of this exercise is to provide digestible meaning to your results upon analysis. Alternatively, you could consider computing a proportion of respondents that "agree/strongly agree." I feel like this result is tangible to readers and meaningful. The mean value alone is susceptible to hiding wide variance in responses.
Thank you so much for a great video, I read a lot about the difference between Likert and Likert type and struggled to understand until I found this video, it is very clear and to the point, Thank you! I subscribed to your channel, and I'm recommending it to everyone I know, its great! I have one question if you don't mind. what references did you use? I need to cite them , again thank you so much
Hi Dal D, very kind comment, thank you! The one reference I cite in that powerpoint is from John Uebersax, which can be found at www.john-uebersax.com/stat/likert.htm You might also consider this paper from a doctoral student.. www.researchgate.net/publication/262011454_Likert
Hi Mussie Mengistu, you can. As long as it meets the criteria of likert scale (not item), you can use those analyses, since they are appropriate for continuous data. See also the question and reply below by Mahim Khan. You will see more info in this video around the 13:30 point of the video. Finally, you may also want to review my (18B) Review of Types of Data video. ruclips.net/video/Mu5uANsEYOE/видео.html
I have a question... We need help on our SIP. How can we determine which/what analysis should we use when we have 3 products with 3 variables to measure... taste, texture and appearance..?. A taste test questionnaire in the form of a likert scale. Ranging from Poor,Fair,Good,Very Good, and Excellent.
If each variable (e.g., taste, texture, and appearance) is measured using a single item (i.e., question), it should be analysed as a categorical variable. if, on the other hand, you have multiple items (generally, 3 or greater) for each variable (or, category--taste, texture, and appearance), then you can analyse those items collectively into a Likert-like variable, which is analysed as a continuous variable.
Hi Joseph, As I say in the video, it is generally considered inappropriate to calculate means of individual Likert items. Instead, you should seek to group multiple Likert items into a Likert scale (see video around 10:00). It's important to ensure there is some validity to your scale (i.e., that the quantitative values and relative differences truly match the reality and you have some evidence of that). Then, once you've done that, you can compute the mean as you would compute any other mean. An example is shown around 13:00 in the video, using the WHOQOL-Brief instrument as an example. That instrument has more detailed instructions in its documentation (can be found via Google search), which you might find helpful. Thank you for your question!
Mahim khan if you'd like to compute z scores you'd need continuous data. so based on what I present here, you could only do that when you've established a likert scale. I wouldn't recommend trying to do it for individual likert items, since I believe it's inappropriate to treat individual likert items as continuous data.
Hello, thanks for very informative and easy to understand video. I have couple of questions and would really appreciate your prompt feedback. Q1: How important is it to have a "Neutral" option in Likert scale. As you have used 4 levels, does it mean it depends on the queries asked? Or is it always preferable to have a "middle" option. The context of this question is that we conducted a questionnaire study in which passengers of a new bus transit system in the city were asked several questions regarding their "level of satisfaction" of various attributes. The actual survey had 5 options: (highly satisfied, adequately satisfied, satisfied, dissatisfied & highly dissatisfied). It does seem that it is skewed more towards "satisfied" option; however having a "neutral" option would also imply that those users are "okay" with the system, thus making them move towards "satisfied group". So, Q2: Should "adequately satisfied and "satisfied" be combined as one category since they essentially tell the same thing. OR should the "satisfied" option be treated as "neutral" since they're essentially saying "we're okay with it"..... Thank you in advance.
A Likert item should be balanced, so in your example, I would personally prefer "highly satisfied," "satisfied," dissatisfied," and "highly dissatisfied." (so 4 options total). This layout is balanced, which is one criteria of a Likert item. Another reason I suggest this option is because I don't see the difference between "adequately satisfied" and "satisfied." They could be considered by some to be equivalent in meaning, thus missing the requirement for Likert items that options are evenly spaced in meaning.
which is the specific test for likert scale data. I have data. for single hypothesis I have a question with 15 subquestions and each subquestion has 5 point likert scale ( Never , rarely , sometimes, frequently, always). In which I can see count for always is very high as compared to other options. I would to do hypothesis test. Mostly non-parametric. So which test would be more useful? Please assist me. Thanks.
Stat Assignment there is no specific test for Likert scale data. It depends on the type of data you are generating (continuous vs categorical) and the sample size/whether the necessary assumptions are met(parametric vs non-parametric). You can view my playlist “Session 19 Inferential Statistics” to learn more about how to choose tests for hypothesis testing and decide based on your own situation
Thank you so much for this video. the difference between a likert scale and likert items is well explained. please do you mind explaining to me what the numbers 6, 5 and 2 stand for or mean in mean.6, mean.5 and mean.2 in the example you gave? another question please: are the scores multiplied by the number of likert scales (4 in the example you gave)? thank you in advance
Thanks for your comments and questions. To help me, can you give the time stamps in the video where you are talking about? I don’t completely understand your question, but if I saw which part of the video you were thinking of I could probably figure it out. Thanks!
Thanks. So, you're looking at the WHOQOL-BREF, which is an instrument from the WHO used to measure quality of life (QOL). It has four domains shown as DOM1-DOM4 in that section of the video. That screenshot comes from the manual for the WHOQOL-BREF. So, when it says mean.6 it is instructing to compute the mean using those values. However, a minimum of 6 values must be recorded for the domain score to be validly computed. You can find the manual at this URL: www.who.int/mental_health/media/en/76.pdf And that information is found on Page 12. The takeaway is that: these instructions are specific to this instrument (WHOQOL-BREF); however, whatever instrument you are using should have instructions, such as the minimum number of responses to be valid. The reason to multiple by 4 in the example given is because the WHOQOL-BREF is a short version of the full instrument (WHOQOL-100). (Multiplying by 4 allows the BREF scores to be comparable to the 100 scores (Can also find that info on Pg 12 of above PDF). So, to answer your question, the multiplication is specific to this instrument as is not something you would necessarily need to do for other instruments.
This video is very beneficial. I was having difficulties labeling the items 'scale' or 'ordinal' since I found conflicting POV's. I was hoping that you attached a reference article about the statisticians' consensus about this topic, that way we can back up our choice.
Hi! Could you tell me do i have to code all items and then recode the reverse items, or can I just go ahead and immediately code the reverse scored ones into reverse?
Hi, it can be done either way. But it’s probably easiest to enter the reverse score directly into the database when doing data entry. That way, any computations or statistics can be directly calculated from the database
@@AnthonyKuster Thank you so much for your quick reply, it was really helpful. Could you please also tell me, if I am trying to find whether lower self esteem levels indicate higher materialism levels, that is, whether they are negatively correlated, which analysis should I use? I am a complete beginner in SPSS and all these options confuse me. In SPSS I have a column of self esteem scores for 1500 individuals, ranging from 1 to 30, and a column of materialism scores, also ranging from 1 to 30. Should I choose for example bivariate -> Pearson correlation, would that be suitable? Thank you in advance for everything.
I forgot to say that the self esteem scores are from a self esteem likert scale and materialism scores are also from a materialism likert scale. So I summed up the answers in order to get the level of materialism and self esteem if each participant, that's how I got my self esteem and materialism score columns.
@@royah5404 Pearson’s correlation should probably work to test whether the two outcomes are correlated. With any statistical test, you have to check assumptions. One thing to look at is a scatter plot. It’s been a long time since I’ve used spss but there are lots of good resources out there. Can search scatter plot spss. Cheers!
@@AnthonyKuster i did this, but there is objection why all the mean scores ranges from 1 to 5 on 5 point likert scale. Is there any cut point or any justification? My mean score is 3.87.
Hello , Sir! Thank you for the video! Can you please give me a reference to support that a Likert scale must be treated as continuous? Thank you again!
Hi, I'm sorry you're having trouble with the video. I just checked the video and didn't see any parts of the video with missing video. Could you please be more specific what the issue is? (what is missing and at what time in the video)? Thank you for your view and comment!
Hey, I saw that some people already posed this question but could you communicate the slides and the literature they're based on? RUclips videos are reasonably no "literature" ;-)
Hi there, I think I have posted a comment in response to a similar question, maybe you can find that one. This material, however, was part of a class I was teaching at the time and was based on my experience along with some material from a textbook/source, which I don’t have on hand (it’s been many years). Hopefully I understand your question correctly and have answered it.
Hi, thanks for the lesson. The example at ruclips.net/video/7m3929CJvcM/видео.html does not have consecutive integers, but you present it as a likert item.
Thanks for pointing that out. The integers aren't shown on the question item; however, in practice, the data glossary would have those values associated with them.
Enormously valuable explanation! Thanks Anthony! And I'm glad to see that you have made lots of other (undoubtedly) presentations! I look forward to learn from them.
AWESOME PRESENTATION, MAY GOD BLESS YOU
Thank you for posting this video, it's very helpful for my thesis.
Thank you for this great lesson!
Thank you so much for the video, it is very illustrative. Big up!
I am the 520th one to have subscribed you, haha, it is a lucky number, great videos!
very informative, and comprehensive, thanks a lot
great illustration
thanks for the brief and clear presentation
Breakthru on my confusion. Thank you.
This was really helpful!!
best explanation
The arrangement can be vertical too. Not restricted to horizontal
Thank you so much. this is rally helpful
thanks bro! helpful!
Hello Dr. Kuster, Great video. Can we email you to ask specific questions?Thank you,
Hi Anthony! Good video! Thank you!
Have you got a literature reference for data analysis methods I can use for my thesis?
did you find this
Very nice explanation sir
Thank you for sharing! It really really helps! Easy to understand and follow your explanation :D
Thank you very much, I appreciate the comment. Please let me know if there’s any other topic you’re interested in.
Hi . thanks for that video. have a question: on 11:37. what are the statistical test or analysis we may run to find the correlation between these 4 domains of health satisfaction? suppose 3 independants variables and 1 dependant variable as they have been measured by likert scale
Hi, thanks for your question. The WHOQOL-BREF produces four Likert scales (the 4 domains of QOL). If you wanted to know the correlation or covariance between these four domains, you would likely use Pearson's correlation coefficient (r). However, if the sample size is small, you would not meet the required assumptions for Pearson's, so you could use non-parametric correlation tests, such as Spearman's rho or Kendall's tau.
Thank you !!
Excellent explanation. Thank you soo much
Thanks so much for your kind words
Hi @Anthony, thank you so much for this clear explanation. Could you please indicate me what video of your channel follows this one? At minute 15:06 you mention that session 19 will deep on this topic. thank you in advance
Hi there and thank you for your interest!
On the channel if you go to playlists you should be able to find
Session 19: Inferential Statistics
Which has 8 videos on topics like Z score, normal distribution, t tests, etc.
Hi Anthony,
Thanks very much for this helpful lecture. I am a student and do not have much experience with data-analysis and SPSS. I hope you can help me with the following question.
I am using a Likert subscale with 3 questions with values 1: strongly disagree up until 5: strongly agree. What I would like to do in SPSS is to compute the mean values of these questions and label them as ordinal variables. My teacher said that that is an option if you do not have a lot of values. Possible values would be then: 1, 1.5, 2, 2.5, 3, 3.5, 4, 4.5 and 5. What would be logical names to use for these computed values?
Thanks in advance, Pieternel
Hi Pieter, it's an interesting question. I've never encountered that approach, so it's hard for me to comment. In terms of creating logical labels, I suspect you would need to do some research to see what people would associate with a term that is "4.5." In other words, what do people generally consider a level of agreement that is part way between "agree" and "strongly agree." Thinking of how I would answer that question myself, I cannot come to an answer. I'd be curious what you decided. I'm assuming the purpose of this exercise is to provide digestible meaning to your results upon analysis. Alternatively, you could consider computing a proportion of respondents that "agree/strongly agree." I feel like this result is tangible to readers and meaningful. The mean value alone is susceptible to hiding wide variance in responses.
Thank you so much for a great video, I read a lot about the difference between Likert and Likert type and struggled to understand until I found this video, it is very clear and to the point, Thank you! I subscribed to your channel, and I'm recommending it to everyone I know, its great! I have one question if you don't mind. what references did you use? I need to cite them , again thank you so much
Hi Dal D, very kind comment, thank you! The one reference I cite in that powerpoint is from John Uebersax, which can be found at www.john-uebersax.com/stat/likert.htm
You might also consider this paper from a doctoral student.. www.researchgate.net/publication/262011454_Likert
can u please suggest? I have a hypothesis h1: elderly people choose social media based on functionality and security . which test should be applied?
thank you
Hi can I use the t-teset, Pearson correlation and regression with the likert scale
Hi Mussie Mengistu, you can. As long as it meets the criteria of likert scale (not item), you can use those analyses, since they are appropriate for continuous data. See also the question and reply below by Mahim Khan. You will see more info in this video around the 13:30 point of the video. Finally, you may also want to review my (18B) Review of Types of Data video. ruclips.net/video/Mu5uANsEYOE/видео.html
I have a question... We need help on our SIP.
How can we determine which/what analysis should we use when we have 3 products with 3 variables to measure... taste, texture and appearance..?. A taste test questionnaire in the form of a likert scale. Ranging from Poor,Fair,Good,Very Good, and Excellent.
If each variable (e.g., taste, texture, and appearance) is measured using a single item (i.e., question), it should be analysed as a categorical variable. if, on the other hand, you have multiple items (generally, 3 or greater) for each variable (or, category--taste, texture, and appearance), then you can analyse those items collectively into a Likert-like variable, which is analysed as a continuous variable.
Thank you sir, but I want to ask a question.
Please how can I calculate mean of likert items using spss?
Hi Joseph,
As I say in the video, it is generally considered inappropriate to calculate means of individual Likert items. Instead, you should seek to group multiple Likert items into a Likert scale (see video around 10:00). It's important to ensure there is some validity to your scale (i.e., that the quantitative values and relative differences truly match the reality and you have some evidence of that). Then, once you've done that, you can compute the mean as you would compute any other mean. An example is shown around 13:00 in the video, using the WHOQOL-Brief instrument as an example. That instrument has more detailed instructions in its documentation (can be found via Google search), which you might find helpful. Thank you for your question!
can we use likert data and analyse the data by using z test in MS Excel ?
Mahim khan if you'd like to compute z scores you'd need continuous data. so based on what I present here, you could only do that when you've established a likert scale. I wouldn't recommend trying to do it for individual likert items, since I believe it's inappropriate to treat individual likert items as continuous data.
Hello, thanks for very informative and easy to understand video.
I have couple of questions and would really appreciate your prompt feedback. Q1: How important is it to have a "Neutral" option in Likert scale. As you have used 4 levels, does it mean it depends on the queries asked? Or is it always preferable to have a "middle" option.
The context of this question is that we conducted a questionnaire study in which passengers of a new bus transit system in the city were asked several questions regarding their "level of satisfaction" of various attributes. The actual survey had 5 options: (highly satisfied, adequately satisfied, satisfied, dissatisfied & highly dissatisfied). It does seem that it is skewed more towards "satisfied" option; however having a "neutral" option would also imply that those users are "okay" with the system, thus making them move towards "satisfied group". So, Q2: Should "adequately satisfied and "satisfied" be combined as one category since they essentially tell the same thing. OR should the "satisfied" option be treated as "neutral" since they're essentially saying "we're okay with it".....
Thank you in advance.
A Likert item should be balanced, so in your example, I would personally prefer "highly satisfied," "satisfied," dissatisfied," and "highly dissatisfied." (so 4 options total). This layout is balanced, which is one criteria of a Likert item. Another reason I suggest this option is because I don't see the difference between "adequately satisfied" and "satisfied." They could be considered by some to be equivalent in meaning, thus missing the requirement for Likert items that options are evenly spaced in meaning.
which is the specific test for likert scale data. I have data. for single hypothesis I have a question with 15 subquestions and each subquestion has 5 point likert scale ( Never , rarely , sometimes, frequently, always). In which I can see count for always is very high as compared to other options. I would to do hypothesis test. Mostly non-parametric. So which test would be more useful? Please assist me. Thanks.
Stat Assignment there is no specific test for Likert scale data. It depends on the type of data you are generating (continuous vs categorical) and the sample size/whether the necessary assumptions are met(parametric vs non-parametric). You can view my playlist “Session 19 Inferential Statistics” to learn more about how to choose tests for hypothesis testing and decide based on your own situation
Thank you so much for this video. the difference between a likert scale and likert items is well explained. please do you mind explaining to me what the numbers 6, 5 and 2 stand for or mean in mean.6, mean.5 and mean.2 in the example you gave? another question please: are the scores multiplied by the number of likert scales (4 in the example you gave)? thank you in advance
Thanks for your comments and questions. To help me, can you give the time stamps in the video where you are talking about? I don’t completely understand your question, but if I saw which part of the video you were thinking of I could probably figure it out. Thanks!
from 13:00 to 13:25. thank you
Thanks. So, you're looking at the WHOQOL-BREF, which is an instrument from the WHO used to measure quality of life (QOL). It has four domains shown as DOM1-DOM4 in that section of the video. That screenshot comes from the manual for the WHOQOL-BREF. So, when it says mean.6 it is instructing to compute the mean using those values. However, a minimum of 6 values must be recorded for the domain score to be validly computed. You can find the manual at this URL: www.who.int/mental_health/media/en/76.pdf
And that information is found on Page 12. The takeaway is that: these instructions are specific to this instrument (WHOQOL-BREF); however, whatever instrument you are using should have instructions, such as the minimum number of responses to be valid.
The reason to multiple by 4 in the example given is because the WHOQOL-BREF is a short version of the full instrument (WHOQOL-100). (Multiplying by 4 allows the BREF scores to be comparable to the 100 scores (Can also find that info on Pg 12 of above PDF). So, to answer your question, the multiplication is specific to this instrument as is not something you would necessarily need to do for other instruments.
Clear enough. Thank you so much for your interaction and clarification.
This video is very beneficial. I was having difficulties labeling the items 'scale' or 'ordinal' since I found conflicting POV's. I was hoping that you attached a reference article about the statisticians' consensus about this topic, that way we can back up our choice.
Consider John Uebersax's website as a starting point: www.john-uebersax.com/stat/agree.htm
Hi! Could you tell me do i have to code all items and then recode the reverse items, or can I just go ahead and immediately code the reverse scored ones into reverse?
Hi, it can be done either way. But it’s probably easiest to enter the reverse score directly into the database when doing data entry. That way, any computations or statistics can be directly calculated from the database
@@AnthonyKuster Thank you so much for your quick reply, it was really helpful. Could you please also tell me, if I am trying to find whether lower self esteem levels indicate higher materialism levels, that is, whether they are negatively correlated, which analysis should I use? I am a complete beginner in SPSS and all these options confuse me. In SPSS I have a column of self esteem scores for 1500 individuals, ranging from 1 to 30, and a column of materialism scores, also ranging from 1 to 30. Should I choose for example bivariate -> Pearson correlation, would that be suitable?
Thank you in advance for everything.
I forgot to say that the self esteem scores are from a self esteem likert scale and materialism scores are also from a materialism likert scale. So I summed up the answers in order to get the level of materialism and self esteem if each participant, that's how I got my self esteem and materialism score columns.
@@royah5404 Pearson’s correlation should probably work to test whether the two outcomes are correlated. With any statistical test, you have to check assumptions. One thing to look at is a scatter plot. It’s been a long time since I’ve used spss but there are lots of good resources out there. Can search scatter plot spss. Cheers!
Please explain how we analyze average point on likert scale data when calculating mean?
To calculate a mean, simply add up the values and divide by the number of data points.
@@AnthonyKuster i did this, but there is objection why all the mean scores ranges from 1 to 5 on 5 point likert scale. Is there any cut point or any justification? My mean score is 3.87.
Hello , Sir! Thank you for the video! Can you please give me a reference to support that a Likert scale must be treated as continuous? Thank you again!
Thanks! I wonder if you could reference a previous study that has used an instrument that generates a Likert scale (such as WHOQOL)?
I like your explantion but it is impossible to see what is on the video from the middle of your talk. Will you please fix it?
Hi, I'm sorry you're having trouble with the video. I just checked the video and didn't see any parts of the video with missing video. Could you please be more specific what the issue is? (what is missing and at what time in the video)? Thank you for your view and comment!
Omgsh I need a personal class from you . Is there any way you take online classes?
Thank you! I’m working on new material. Are there other topics you’re interested in?
Hey,
I saw that some people already posed this question but could you communicate the slides and the literature they're based on? RUclips videos are reasonably no "literature" ;-)
Hi there, I think I have posted a comment in response to a similar question, maybe you can find that one. This material, however, was part of a class I was teaching at the time and was based on my experience along with some material from a textbook/source, which I don’t have on hand (it’s been many years). Hopefully I understand your question correctly and have answered it.
How can we calculate the total knowledge score?
Thanks for your question, what are you trying to calculate?
❤
how to input likert scale data in minitab? your response would be a really great help for my research 🥺 btw, you have the best explanation 😊
Hi thanks for your compliment. Unfortunately I’m not familiar with minitab. Only stata, excel
@@AnthonyKuster thankyouu for your response, how about in excel? may I know how to input likert scale datas to solve for its correlation?
@@airabellem.3939 hi, I have a similar problem! Did you manage to get the answer you wanted? Thank you
how can i quote you in my thesis
Thanks for your kind comment. Most of what I discuss is cited in slides in the video. Not sure RUclips is acceptable for thesis citation... yet 🤓
@@AnthonyKuster you are very informative, thank you!! However when coding and analysing, how should we treat the neutral reponses
Hi, thanks for the lesson. The example at ruclips.net/video/7m3929CJvcM/видео.html does not have consecutive integers, but you present it as a likert item.
Thanks for pointing that out. The integers aren't shown on the question item; however, in practice, the data glossary would have those values associated with them.
very informative, and comprehensive, thanks a lot
Glad it was helpful!
thank you
thank you
My pleasure!