What is the Bonferroni correction?

If you’ve read a few research studies, then you will doubtless have seen the expression “Bonferroni correction” in the statistical analysis section. But what does it mean? Why is it important? In this guest post, Tim Egerton from Sport Science Tutor explains what an important recent review paper found:

Research sometimes involves making multiple comparisons between groups. However, making multiple comparisons increases the chances of making a type I error (the risk of identifying a difference when there is not really a difference). Good statistical treatment of data can reduce this risk. A popular statistical treatment used for this purpose is the Bonferroni correction. When researchers fail to use the Bonferroni correction correctly, it increases the likelihood that their paper has reported an incorrect finding. Therefore, for interpreting research properly, we need to know when it should be used.

The study: When to use the Bonferroni Correction, by Armstrong, in Ophthalmic and Physiological Optics, 2014

-

What is the background?

The Bonferroni correction is commonly used to adjust probability (p) values to correct for a familywise error rate when making multiple comparisons on the same set of subjects. However, the test is so commonly used that it has become routine. And now, the routine use of this test has been criticised as resulting in a reduction in the chance of a type I error occurring at the expense of type II error.

-

What are the key concepts?

The following terms and concepts are key for an understanding of this article. Especially important are type I error and type II error. The definitions are as follows:

Type I error – This is a rejection of the null hypothesis when the null hypothesis is true. In practical terms, this means finding a difference when there isn’t one.

Type II error – This is an acceptance of the null hypothesis when the null hypothesis is false. In practical terms, this means failing to find a difference when there is actually one.

Null hypothesis – This is the hypothesis that there will be no differences between groups in a study.

Familywise error rate – This is an inflation of the error rate when making a series of comparisons on the same set of subjects.

-

What did the researchers do?

The researchers assessed the extent to which statistical tests, in particular the Bonferroni correction, are used to reduce the chance of a type I error in research studies. They were also interested in whether researchers routinely provide a rationale for their test selection. They did this by reviewing current practice in the use of Bonferroni and other types of correction in three journals over a 10-year period. Two searches were made to assess:

The frequency of correction of p values by any available method (Search terms: ‘multiple testing’, ‘post-hoc’ tests). The question the researchers were interested in was: did the article correct p values to reduce the chance of a type I error using any of the available methods and provide a rationale for the method used?

The specific use of the Bonferroni adjustment (Search terms: ‘Bonferroni correction’, ‘Bonferroni adjustment’, ‘Bonferroni post-hoc test’). The question the researchers were interested in was: did the study apply Bonferroni correctly and did it provide an appropriate rationale and/or discussion of its use?

- 

What happened?

Frequency of correction of p values

Of the 187 studies reviewed, the reviewers found that:

142 articles included multiple statistical testing, of which 95 (67%) corrected p values and 47 (33%) did not correct p values when performing multiple comparisons.

Of the 95 articles that did correct p values, only 9 provided a clear rationale for the correction (to avoid a type I error), while 86 provided no clear rationale.

-

Use of Bonferroni adjustment

Of the studies reviewed, the reviewers found that:

187 used the Bonferroni correction, of which 133 (71%) provided no discussion on the rationale or discussion of the method.

Of the 54 of the articles that provided some discussion, 36 considered its relevance in reducing a type I error, 2 discussed the possibility of a type II error, 6 discussed the relevance of the Bonferroni correction and decided not to adjust p values, and 8 gave an incorrect rationale for its use.

-  

What did the researchers conclude?

The researchers concluded that:

The Bonferroni correction is indeed the most popular statistical test for correcting p values in studies making multiple comparisons.

The majority of studies do not consider the relative risks of type I and type II errors in relation to the use of the Bonferroni correction.

A significant number of articles do not provide any rationale or discussion of the method of correction used for multiple comparisons or its consequences.

-

What are the limitations?

The researchers provided a number of recommendations for the appropriate statistical treatment of data when multiple comparisons are involved in a study. However, these recommendations were not based upon the findings from the review.

-

What are the practical implications?

The appropriate use of the Bonferroni Correction appears to depend on the relative importance of avoiding a type I error or type II error in research involving multiple comparisons.

Many researchers appear to use the Bonferroni Correction without due consideration of the appropriateness of this test.

When reading research, care should be taken to appraise the statistical treatment. Inappropriate statistical treatment may cause type I or type II error, resulting in incorrect conclusions being made.

If you are a student and need help with your statistics or methods, or if you just want someone to help you with an assignment, don’t miss Tim’s special offer for readers of S&C Research. Click HERE to learn more!

Sport Science Tutor

FREE NEWSLETTER: which type of periodization is best?

Tomorrow, the October edition of the free newsletter goes out. It covers a paper in which a group of researchers performed a systematic review and meta-analysis into which types of periodization are best for strength gains.

Are you curious to find out whether daily undulating periodization is as effective as some strength coaches claim it to be? What do you think the researchers found? Just subscribe to the free newsletter to find out!

To get the full study review when it is sent out tomorrow, just enter your email below. This study review will not be published on the blog and it will not appear in the monthly research review. So don’t miss out. Sign up now!

What is the fiber type of different muscle groups?

Training in line with the muscle fiber type of body parts could be useful for maximizing hypertrophy. However, for too long, information about muscle fiber types has been promoted based on references to single studies and hearsay. In this article, Chris Beardsley reviews the literature regarding muscle fiber types for the main upper and lower body muscles.

What is the background?

What are muscle fiber types?

Muscle fibers can be classified in various ways. All current methods are dependent upon the assumption that limiting factor for the speed at which cross-bridge cycling can occur is the speed at which the ATPase of the myosin head can hydrolyze ATP to power the process. The three ways are:

Myosin ATPase histochemical staining – this process differentiates between individual muscle fibers on the basis of their staining intensities. These staining intensities differ between muscle fibers as a result of differences in pH sensitivity. The differences indirectly provide relative information between muscle fibers about the speed at which ATP hydrolysis occurs, although the staining procedure does not directly measure the speed of hydrolysis (Scott, 2001). The main three myosin ATPase staining results are referred to as muscle fiber types I, IIA, and IIX (or historically IIB), respectively. Interim fiber types are identified where staining types between the main classes are observed.

MHC isoform identification – this process involves differentiating between individual muscle fibers in the basis of the different myosin heavy chain isoforms. The MHCs contain the site that serves as the ATPase, which is how identifying the MHC isoform is relevant for the speed of ATP hydrolysis. Each muscle fiber can contain more than one MHC isoform. Thus, although there are only three isoforms expressed in human skeletal muscle, there are many more hybrid muscle fiber types comprising muscle fibers with several different isoforms in the same muscle fiber (Scott, 2001). The main three myosin isoforms are most correctly referred to as MHCI, MHCIIa, and MHCIIx (or MHCIIb historically).

Biochemical identification of metabolic enzymes – this process combines information derived from myosin ATPase histochemistry with histochemistry of certain enzymes that are involved in energy metabolism. In such cases, the myosin ATPase fiber typing is used to classify muscle fibers into either type I or type I. Analysis of enzymes is then performed in order to provide information about the metabolic pathways. This leads to describing the muscle fibers as either aerobic/oxidative or anaerobic/glycolytic and ultimately three different fiber types: fast-twitch glycolytic, fast-twitch oxidative, and slow-twitch oxidative (Scott, 2001).

In many circles, the three different methods are taken as producing similar outputs that can be compared across studies. While this has been found to be acceptable for certain muscle fiber types and between certain typing methods (most obviously in respect of type I muscle fibers and between MHC and myosin ATPase), it is certainly not valid across the board and we must bear this in mind when reviewing the literature.

-

Why are muscle fiber types relevant?

Type II muscle fibers have greater growth potential than type I muscle fibers. This has led most experts to recommend focusing on type II muscle fiber types in resistance-training for hypertrophy. However, as Ogborn (2014) has explained, type I muscle fibers can also be developed in order to maximize the overall increase in size of a muscle. Thus, hypertrophy programs should be structured with both muscle fiber types in mind.

-

How can different muscle fiber types be developed?

Type I (often also called slow twitch) fibers are highly resistant to fatigue and therefore probably respond best to sets of higher repetitions. On the other hand, type II (often also called fast twitch) fibers fatigue quickly and probably respond best to high-load, low-repetition sets. Mixtures of type I and type II muscle fibers will probably need a combined approach, perhaps through a combination of high repetitions and others to low repetitions.

-

Why is the muscle fiber type of individual muscles useful?

As we will see in the following analysis, some muscles contain a greater proportion of type I muscle fibers while others contain a higher proportion of type II muscle fibers. Where a muscle does contain a higher proportion of type I muscle fibers, it seems probable (although it remains to be shown in long-term trials), that training with sets of higher repetitions to muscular failure might well lead to superior muscle growth, at least in the short-term. Similarly, where a muscle contains a higher proportion of type II muscle fibers, it seems probable that training with high-load, low-repetition sets might well lead to superior muscle growth, at least in the short-term.

***

What are the muscle fiber types of the lower body muscles?

The studies in the tables below have explored the muscle fiber types of the lower body musculature:

Screen Shot 2014-09-05 at 14.36.51In summary, the hamstrings and gluteus maximus comprise a mixed-to-slow muscle fiber type. The soleus is predominantly slow twitch. The gastrocnemius possesses a mixed-to-slow muscle fiber type. The mixed-to-slow fiber type gastrocnemius is a two-joint muscle while the very-slow twitch soleus is only a single-joint muscle. By sitting down to perform plantar flexion, the gastrocnemius enters active insufficiency and the soleus is primarily recruited. In contrast, during standing exercises, the gastrocnemius is more involved. So seated plantar flexion exercises may benefit most from a focus on type I muscle fibers.

Lower body muscle fiber typesThe individual quadriceps range from mixed-to-slow through to fast twitch. The rectus femoris is a two-joint muscle and this is fast twitch. Thus, multi-joint knee extension exercises may benefit more from a focus on type II muscle fibers than single-joint knee extension exercises.

***

What are the muscle fiber types of the upper body muscles?

The studies in the tables below have explored the muscle fiber types of the upper body musculature:

Upper body muscle fiber typesThe shoulders are mixed-to-slow twitch muscle fibers, while the biceps, triceps and pectorals are mixed-fast twitch. Most studies have reported that the latissimus dorsi is almost perfectly balanced between slow and fast twitch muscle fibers. A more recent investigation cast doubt upon this using a new method. Future research may therefore find that there is a greater proportion of fast twitch fibers. However, this remains to be seen.

***

What do we know about these studies?

If you’re curious about the methodology used in the studies cited above (i.e. whether the researchers used histochemical analysis or immunohistochemical methods, and whether they took biopsies from living subjects or samples from cadavers) or if you simply just want the references, here are the details and links. They’re in alphabetical order:

Baker and Hardy (1989) took muscle samples from the latissimus dorsi by needle biopsy, at rest, before and after a high-intensity exercise training canoe-specific training program, performed 3 times per week for 9 weeks.

Dahmane (2000) explored the histochemical properties of several muscles in two groups of 15 men aged 17 – 40 years using the histochemical analysis of myosin ATPase. The muscles investigated included the biceps brachii, triceps brachii, flexor digitorum superficialis, extensor digitorum, biceps femoris, tibialis anterior and gastrocnemius. They also tested a tensiomyographic, non-invasive measurement technique.

Dahmane (2005) investigated fiber type distribution in 9 limb muscles with histochemical methods in two groups of 15 men aged 17 – 40 years at different depths (superficial and deep).

Dahmane (2006) measured muscle fiber type proportions in samples of the biceps femoris in sedentary young men using the standard histochemical analysis of myosin ATPase as well as the MHC isoform expression test using immunohistochemical analysis.

Edgerton (1975) explored the muscle fiber types of the gastrocnemius, soleus, vastus lateralis and intermedius in 32 humans by autopsy within 25 hours of death. The samples were examined using histochemical analysis of myosin ATPase.

Garrett (1984) explored the muscle fiber type composition of the human hamstring muscles using histochemical analysis of myosin ATPase of necropsy specimens taken from seven locations in the hamstring.

Gouzi (2013) systematically reviewed the studies providing data on fiber type proportion of the vastus lateralis of the quadriceps in healthy subjects aged >40 years old. The methods used in the underlying studies differed and included both histochemical analysis of myosin ATPase and immunohistochemical analysis of MHC isoforms.

Hards (1990) examined the muscle fiber type of internal intercostal, external intercostal, and latissimus dorsi muscle biopsies in 68 patients who were having a thoracotomy using myosin ATPase.

Humphrey (1982) performed an investigation into the muscle fiber composition of the deltoid muscle of elite British slalom kayak competitors for a thesis at the University College of North Wales.

Jennekens (1971) gathered data regarding the distribution of muscle fiber types in 5 human limb muscles using necropsy material from 8 previously normal subjects aged from 8 – 42 years, who died suddenly as a result of trauma or acute illness. The muscles explored included the deltoid, biceps brachii, rectus femoris, gastrocnemius and extensor digitorum brevis. The researchers used histochemical analysis of myosin ATPase to assess muscle fiber type.

Keh-Evans (2010) investigated the contractile, histochemical and biochemical properties of the triceps surae were compared in 13 aerobically-trained male subjects aged 63 – 76 years.

Johnson (1973) also gathered data on the distribution of muscle fiber types in 36 human muscles using necropsy material from 50 sites of 6 previously normal male autopsy subjects aged from 17 – 30 years. The researchers used histochemical analysis of myosin ATPase to assess muscle fiber type.

MacDougall (1984) took muscle biopsies and estimated muscle fiber numbers in vivo in biceps brachii in 5 elite male bodybuilders, 7 intermediate caliber bodybuilders, and 13 age-matched controls. The researchers used histochemical analysis of myosin ATPase to assess muscle fiber type.

Mavidis (2007) took muscle biopsies of the deltoid muscles of Greek professional male tennis players. Myosin ATPase histochemistry and MHC composition analysis were performed on the samples.

Nygaard (1982) investigated the muscle fiber type at nine sites of the brachial biceps and the lateral vastus at autopsy from 5 elderly subjects.

Pierrynowski (1985) evaluated hamstring muscle fiber type in order to develop a musculoskeletal model.

Schantz (1983) took muscle biopsies were taken from the vastus lateralis and triceps brachii and used histochemical analysis of myosin ATPase to assess muscle fiber type.

S̆irca (1980) assessed muscle fiber type of the gluteus maximus, gluteus medius and tensor fasciae latae in patients with osteoarthritis of the hip as well as in autopsy material.

Srinivasan (2007) performed a study of the muscle fiber type composition of 14 muscles spanning the glenohumeral joint. They took material from 4 male cadavers (mean age 50 years) within 24 hours of death and performed histochemical analysis of myosin ATPase to assess muscle fiber type.

Tesch (1983) took muscle biopsies from the mid-portion of deltoid muscles of 7 male wheelchair basketball athletes, 8 high caliber kayak paddlers, 8 wrestlers, and 8 mountain ranger soldiers. performed They used histochemical analysis of myosin ATPase to assess muscle fiber type.

Travnik (1995) took cross-sections from autopsied muscles from 9 healthy men, aged 18 – 44 years, who had died suddenly. They used histochemical analysis of myosin ATPase to assess muscle fiber type.

N.B. This list is not exhaustive in all respects. In particular, if you go digging, you may come across a number of other studies that have explored muscle fiber types in either the vastus lateralis or the gastrocnemius. Unlike other parts of the body, these muscles have been analysed in many investigations. While muscle fiber type does vary, the prevailing trend (as reported in review papers) is for the fiber type in the vastus lateralis to be mixed (around 50%). While the gastrocnemius has not yet been subjected to a review, it does seem to display a mixed-to-slow twitch fiber type.

***

What are the practical implications?

For the lower body

The hamstrings and gluteus maximus comprise a mixed-to-slow muscle fiber type. Hip extension exercises, as well as knee flexion exercises, may benefit from being performed with a combined approach with an emphasis on higher repetitions to muscular failure.

The soleus is predominantly slow twitch. The gastrocnemius possesses a mixed-to-slow muscle fiber type. Since sitting preferentially loads the soleus, seated calf raises may benefit from being performed exclusively with higher repetitions to muscular failure, while those performed standing may also benefit from a minor amount of high-load training.

The individual quadriceps range from mixed-to-slow through to fast twitch muscle fiber types. The rectus femoris is a two-joint muscle and is fast twitch. Thus, multi-joint knee extension exercises like squats may be best performed for high-load, low-repetition sets while single-joint knee extension exercises may benefit from a combined approach to loading.

For the upper body

The shoulders are mixed-to-slow twitch muscle fibers, suggesting that they may benefit from being subjected to a combined approach with an emphasis on higher repetitions to muscular failure.

The biceps, triceps and pectorals are all mixed-fast twitch, suggesting that they should be best trained with a combined approach with an emphasis on high-load, low-repetition sets.

The latissimus dorsi is almost perfectly balanced between slow and fast twitch muscle fibers, suggesting that it should be best trained with a combined approach.

FREE NEWSLETTER: does dieting change brain activity?

Tomorrow, the September edition of the free newsletter goes out. It covers a paper in which reviewers explored recent research into brain responses to food in healthy and obese individuals, as well as in response to periods of weight loss.

Studying the brain responses to food is challenging. However, functional magnetic resonance imaging (fMRI) and positron emission tomography (PET) are tools that have been recently developed for appetite research. These tools have allowed researchers to deduce some important findings.

What did the reviewers find? Well,you’ll need to read the newsletter to find out!

To get the full review when it is sent out tomorrow, enter your email below. This review will not be published on the blog and it will not appear in the monthly research review. So don’t miss out!

« Older Entries