The dullest post in this blog: Exploratory Factor Analysis and how on Earth is it related to music psychology

When the lonely Jyväskylän summer hits you inside your untidy studio, your mind dangerously starts wandering into regions which you have never contemplated before: learning more Finnish, raising a plant and calling it “Bob”, reading Camus’ The Pest, and, of course, starting to read a Confirmatory Factor Analysis book. Dear readers, this is the beginning of, by far, the dullest post in this MMT blog. I dare you to bear with me through this hopeless text.

la peste
The best companion for a kind of sunny Finnish summer.

CFA imageI recently came across with Timothy A. Brown’s Confirmatory Factor Analysis for Applied Research book (2006). We could include this title among the field of psychometrics. This brunch of psychology deals with (among many other things) analyzing questionnaires of several kinds which are used by researchers across disciplines. Just as an example from my own field, think about famous psychological inventories or tests which you might have come across at some point in your life. Does the Big 5, the Wechsler Intelligence Scale or the Beck Depression Inventory rings a bell? Well, it turns out that creating these tests is by itself a titanic endeavor which demands a lot of time and brains before researchers can happily survey hundreds of people and draw conclusions in their all mighty publications.

So, before you close this tab and return to watch funny hedgehog videos in YouTube (here is one), let me introduce the topic of this post. Here it goes: Exploratory Factor Analysis or EFA. Even though Brown’s book deals mainly with Confirmatory Factor Analysis (or CFA from now on), let’s put it aside for a later post (yes, it is a threat) and focus for now on a brief description and explanation of EFA. Let’s tackle this concept with a silly example. Let’s suppose that a group of Martians arrive to the Earth. They know nothing about what we eat, so they go to your local supermarket to find it out. After collecting as much samples of food as possible they place everything in the ground and stare at it.

food
What the Martians found…

At first, nothing really makes sense: it just looks like a bunch of random terrestrial food. But then, some of them start grouping the food according to shared characteristics. They figure out, for instance, that different pink, red, and beige things share certain softness, bad smell and some red liquids. They also create another group made of things that seem fresh, juicy, and which contains seeds. After repeating this process some more times they come up with different sets of food, each of which hold an underlying characteristic which is present despite the subtle difference within each group. Furthermore, they come up with names for each set which correspond to our classification of meat, vegetables, fruits, carbohydrates, etc. We can all agree that there are many kinds of meat and they all have differences (i.e. they all represent different manifestations of the concept “meat”, despite their acknowledgeable differences) but, they all share necessary characteristics which enables their belonging to the label “meat”. The Martians accomplished a process of identifying shared characteristics between different kinds of food and finding a smaller number of categories to group them into.

food piradmid
…what the Martians ended up with.

Well, EFA can broadly be understood as the food-sorting process that the Martians undertook. EFA is a technique which needs as input a collection of elements (e.g. indicators, variables, items of a questionnaire) which are related to each other in certain ways and then through crazily complicated statistical procedures comes up with a smaller number of invisible factors (e.g. the labels “meat”, “vegetables”, etc.). These can account for the observed relationships between the collections of elements as best as possible, while mathematically recognizing that there are measurement error in the collected elements (each food within a label have their own differential characteristics).

The statistical and conceptual aspects of EFA are way too complicated to be explained in this space (the truth is that I can hardly understand them and let alone come up with other silly examples for them as well), but if anyone is interested in finding out more about this curious statistical technique, I would definitely suggest reading Brown’s book (particularly chapter 2) and the excellent paper by Preacher and MacCallum (2003). Besides describing the EFA process and requisites step by step, they make a crystal clear (well, kind of) differentiation between EFA and Principal Component Analysis, another data reducing technique which often is mistaken as EFA (they are not the same!).

If you reached this far you might be asking “what on Earth does this has to do with music psychology?” Well, even more than what you think! In music psychology research, we often make use of different techniques to collect data. One popular way of doing it is through psychological self-report questionnaires (some of which I mentioned before). I want to use one questionnaire as an illustration: Dianna T. Kenny’s Music Performance Anxiety Inventory (K-MPAI, 2009a). However, before that we need to define “music performance anxiety”.

Music performance anxiety (MPA) is

The experience of marked and persistent anxious apprehension related to musical performance that has arisen through underlying biological and/or psychological vulnerabilities and/or specific anxiety-conditioning experiences. It is manifested through combinations of affective, cognitive somatic, and behavioral symptoms. It may occur in a range of performance settings, but is usually more severe in settings involving high ego investment, evaluative threat (audience), and fear of failure. It may be focal (i.e. focused only on music performance), or occur comorbidly with other anxiety disorders, in particular social phobia. It affects musicians across the lifespan and is at least partially independent of years of training, practice, and level of musical accomplishment. It may or may not impair the quality of the musical performance (Kenny, 2009b, p. 433).

So, we have quite a complex definition of what people might popularly know as “stage fright” in musical contexts. As expressed above, musicians might have performance impairment due to uncontrollable MPA levels. In fact, MPA can have threatening consequences for the health and career of musicians such as quitting their profession or consuming drugs, among others (Kenny, 2011; Ortiz, 2011a; Taylor & West, 2004; West, 2004). How to help musicians who experience impairing levels of MPA? Good will is not enough and potential solutions are quite complex. However, an important first step towards the goal is to have solid psychological measurement tools. Here is where EFA can prove to be a powerful ally.

The revised K-MPAI (Kenny, 2009a) contains 40 items that are supposed to reflect the complex conceptual structure described before. But one can’t simply take the questionnaire, survey a bunch of musicians, decide who has high or low MPA levels and introduce some sort of treatment. Before that, we need to find evidences that our items are indeed reflecting the MPA concept. In other words, we want to make sure that the different pink, red, and beige things that share certain softness, bad smell and some red liquids (i.e. the 40 items) are indeed related to each other under the category of “meat” (i.e. MPA). How do we do that?

In a very schematized way, we would have to ask musicians to fill K-MPAI questionnaires. Authors differ with regard of how many surveyed-people do we need . But, usually it’s a lot (for a more detailed discussion check Tinsley & Tinsley, 1987). For example, Nunnally (1987) advised collecting 10 people per each item in a questionnaire. Since we have a 40-items long inventory, we would need at least 400 musicians surveyed with the K-MPAI! Only after collecting the right amount of participants, then we would be able to proceed with the EFA (i.e. carry on with the Martian classification method). The next steps required to conduct a proper EFA are selecting the right factor extraction method, then selecting the rotation method, after that, choosing the method for estimating the number of factors to retain, and then the… Ok, you got the point: it’s quite a complicated process filled with satanic-sound-like terms. My point is that conducting an EFA is an elaborated thing to do. But, what do we end up with by the end of the process (if we don’t pull our eyes out during the attempt)?

Check the following list of items from the K-MPAI. They are just a few examples from the whole 40-items inventory. But before you do, remember to think like a Martian. Can you spot some similarities and differences that could help us classify the list into smaller groups?

  • I often feel that I have nothing to look forward to
  • I never know before a concert whether I will perform well
  • Prior to, or during a performance, I experience dry mouth
  • I often feel that I am not worth much as a person
  • During a performance I find myself thinking about whether I’ll even get through it
  • I generally feel in control of my life
  • Even if I work hard in preparation for a performance, I am likely to make mistakes
  • Sometimes I feel depressed without knowing why
  • Prior to, or during a performance, I get feelings akin to panic
  • I often find it difficult to work up the energy to do things

In a previous research with a sample of Peruvian tertiary music students (Chang-Arana, 2015), I conducted one type of EFA (yes, there are different types of them) after of which I ended up with this division:

Group 1:

  • Even if I work hard in preparation for a performance, I am likely to make mistakes
  • Prior to, or during a performance, I get feelings akin to panic
  • I never know before a concert whether I will perform well
  • Prior to, or during a performance, I experience dry mouth
  • During a performance I find myself thinking about whether I’ll even get through it

Group 2:

  • I generally feel in control of my life
  • I often feel that I have nothing to look forward to
  • Sometimes I feel depressed without knowing why
  • I often find it difficult to work up the energy to do things
  • I often feel that I am not worth much as a person

Group 1 seems to have specific concerns regarding performance, and Group 2 seems to be more related to depressive symptoms. We have created two categories or factors, just like the Martians used “meat” or “vegetables” for classifying the terrestrial food.

So, how can this help us to help our anxious musicians? Well, we obtained some evidences that when a participant scores high in, let’s say, “I never know before a concert whether I will perform well” it gives us a measure on specific concerns regarding performance. Similarly, if the same person scores low in “I often feel that I am not worth much as a person”, then we have a measure on a person’s depressive symptoms. In a very simplified way, we could say that the more items we have had correctly classified for each factor, the more angles of MPA we can stare at, and the more informed decisions we can make.

Nevertheless, EFA is not the holy grail of questionnaire testing. In fact, many much other steps need to be taken in order to assess the appropriateness of using a questionnaire. Each time we want to use a questionnaire out of the context in which it was originally reported, we need to assess it again. If we change the cultural context (let’s say, evaluating musicians from different countries), some characteristics might be shared, but others will be very different. They need to be identified before we can make decisions based on the scores of whatever questionnaire we happen to be working with (AERA, APA, & NCME, 2014). EFA, although powerful, is just one more way of achieving this goal.

HolyGrail034
Exploratory factor analysis is not this.

To conclude, I have attempted to explain the main concept behind a fairly popular statistical technique: factor analysis and, particularly, exploratory factor analysis. I raised some silly examples and attempts of jokes in order to illustrate its worth in research and particularly in music psychology research. In order to do so, I made use of music performance anxiety literature and one of the inventories created for measuring it. Through it, we saw how EFA can help us to reduce an otherwise large amount of information (items) into a smaller meaningful sets of groups (factors). I hope that the peaceful Finnish summer gave me some sort of zen-mind set to make this post fluid and readable. Regardless, I’ll go back to The Pest and keep questioning the meaning of life under the endless Finnish summer days.

IMG_1571

 

References

American Educational Research Association, American Psychological Association, & National Council on Measurement in Education (2014). Standards for educational and psychological testing. Washington D.C.: American Educational Research Association.

Brown, T. A. (2006). Confirmatory factor analysis for applied research. USA: The Guildford Press.

Chang-Arana, A. M. (2015b). Adaptation and psychometric properties of the Kenny-Music Performance Anxiety Inventory (K-MPAI) (Unpublished license thesis). University of Lima: Peru. doi:10.13140/RG.2.2.14697.49763

Kenny, D. T. (December, 2009). The factor structure of the revised Kenny Music Performance Anxiety Inventory. Research presented at the International Symposium on Performance Science, Auckland, New Zealand. Retrieved from http://www.legacyweb.rcm.ac.uk/cache/fl0019647.pdf

Kenny. D. T. (2009b). Negative emotions in music making: Performance anxiety. In P. Juslin & J. Sloboda (Eds.), Handbook of music and emotion: Theory, research, applications. Oxford, UK: Oxford University Press.

Kenny, D. T. (2011). The psychology of music performance anxiety. Oxford: Oxford University Press.

Nunnally, J. C. (1987). Teoría psicométrica [Psychometric theory]. México D.F.: Trillas.

Ortiz, A. (2011a). Music performance anxiety-Part 1: A review of its epidemiology. Medical Problems of Performing Artists, 26(2), 102-105.

Preacher, K. J., & MacCallum, R. C. (2003). Repairing Tom Swift’s electric factor analysis machine. Understanding Statistics: Statistical Issues in Psychology, Education, and the Social Sciences, 2(1), 13–43. http://dx.doi.org/10.1207/S15328031US0201_02

Taylor, A., & Wasley, D. (2004). Physical fitness. In A. Williamon (Ed.), Musical excellence: Strategies and techniques to enhance performance (pp. 163-178). New York: Oxford University Press.

Tinsley, H. E. A., & Tinsley, D. J. (1987). Uses of factor analysis in counseling psychology research. Journal of Counselling Psychology, 34(4), 414-424.

West, R. (2004). Drugs and Musical Performance. In A. Williamon (Ed.). Musical Excellence: Strategies and Techniques to Enhance Performance (pp. 271-290). New York: Oxford University Press.

Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s