January. Happy New Year. FREE Internet for Smart/Globe Even with Zero Load for Android and for PC/Laptop.
*Visit Coolchura for news, trending topics, discussions and promos.
*Interested in blogging and making-money online? Try Colored Chalks
*Utilize the
Search Blog Contents if you are looking for certain keywords. It really works! You can also try the Ctrl+F for more convenience.
Problem downloading? Use mirror links or click here for instructions.
Please use the comment form to report nonworking download links.

Thursday, March 5, 2009

ANOVA and Pearson for Likert Scale Analysis, Huh?

I had been analyzing data of masteral students (my teacher corrected me before that there is no "masteral" word but "master's student" doesn't sound good). One problem I had encountered was the use of parametric tests to those that concern ordinal variables. My professors in UP taught me that Always, Often, Sometimes and Never are ordinal variables and they had also quoted that the measurement of this likert scale is still debatable to whether it is nominal or ordinal. I analyzed two questionnaires with this type of measurement using nonparametric tests like Mann-Whitney, McNemar, Friedman Test and Kruskal-Wallis but the panelist who checked them both refused the tests that I had done. They suggested the ANOVA and t-Test for the differences and reconstructed a problem to concern correlation/association. The tests they suggested for the correlation was Pearson R. What?!

My classmate once asked my professor this issue on ANOVA for the likert scale and yes they agreed that the test is not appropriate because ANOVA is for interval variables. So, Pearson R is also not an appropriate test for the correlation of ordinal variables as for the rule. There are no special rules for the use of ANOVA and Pearson for the ordinal variables, right? I just followed the panelists’ advice because I want those students to pass their thesis defense. Even if I feel that I violated my learning, I am still not in the position to contest their knowledge because they had been advising theses before I was able to gain my statistical knowledge. It’s like they are veterans already and I am just a fresh bud. It just confirmed my belief that some schools in my country are not providing the students with the right learning. I hope that those in their level will be kind enough to tap these people's backs and enlighten them on the issue. I may be wrong with this so please let me know!

Like this post? Subscribe or Bookmark my posts!


  1. This creates a problem for me now.
    I have just completed a survey with a questionnaire mainly consisting of Likert-scale-type questions. I used Pearsons r to analyze the data for predicted correlations because it was the easiest thing to do. In literature many say that this is wrong, but there is no common few of what would be the correct measurement. And then there seem to be schools that even say you have to use Pearsons R. Very confusing!
    What I have done to resolve the problem is to print scatter plots for predicted correlations to get a visual idea of what is going on. I hope it's sufficient to get my thesis through.


  2. i am working with my thesis now i felt confused about pearson r or any of those statistical treatment for my hypothesis or for the data collected in my questionaire. i am no expert of numbers nor statstical method but i need help.

  3. You just have to ask your thesis adviser. There are also schools who require students to have a statistician as part of the panel. You can also consult from them.


About the Author

My photo
My field is in research and analysis. And it is to my experience that having answers in a bunch is a rare find. That's why I started blogging. To provide answers in one gulp as much as possible. Follow my posts by clicking the RSS Feed. Your donation is highly appreciated. If you want to exchange links, contact me or use the comment form.
Visit my personal page

  © Blogger template Spain by Ourblogtemplates.com 2008

Back to TOP  

Related Posts Plugin for WordPress, Blogger...