The anti-fat police are really getting quite desperate, if they think people will take this crap seriously. Lets analyse this a teensy bit, shall we? Firstly, let us ignore the actual test and results and look first at the group that was tested, bit by bit. 9-13 year old "obese" girls already taking part in a weight loss group.
I started puberty at the age of 10, approximately. Most young girls start puberty at some point between the ages of 9 and 13, and with the onset of puberty comes
- Growth spurts in height
- Accumulation of breast tissue
- Shifting and increasing fat deposits at hips, buttocks, stomach and breasts
- Sweling and growth of the genitals
- The monthly period cycle, including the regular peaks and troughs every month in weight, hormones, period blood and water retention
- Fluctuations in weight against height as the body settles into its new shape whilst trying to grow really, really fast
On this matter alone we can see that certain fluctuations in BMI are to be expected in the 9-13 age-group, and as a result any experiment testing the BMI changes in such children should account for this in both the conclusions and study, and include an additional chunk in their margin of error for any calculations. This still may not be reliable, however, in smaller test groups, as there really is no set amount of weight, height or BMI that a child is "supposed" to gain during the onset of puberty, especially considering late and early bloomers.
Having considered the above, let us now take a closer look at the experiment itself. Any good scientist knows how to be rigorous, and that certain things can be done to increase the potential accuracy of any results. Some of these are as follows.
In smaller groups, a single outlier can have a statistically large affect on the overall result, skewing it unfairly. As a result, larger study groups are best as the results from a single individual will have a smaller overall effect, verging to the point of being negligible, as the potential difference becomes significantly smaller than the calculated margin of error. It is also best to take a reasonably specific group. "Apples" being tested for average mass and colour may vary from the massive, pale green cooking apples to tiny, pink-hued ones so popular for childrens lunch boxes. A test study that contains mostly small, pink apples would be significantly skewed in results if a large cooking apple is included.
More measurements is always better, excepting the act of measuring itself may affect the results under certain circumstances (Schrodinger's Cat being an obvious example). If I want to calculate the speed of light, and whether it is significantly affected by travelling through different media (air, vacuum, glass, etc), I would be a very poor scientist indeed to measure once, in each medium, and then publish my results as conclusive. The equipment may have malfunctioned, or my results may be skewed by the margin of error, or there may have been a fault in one or more of the media tested that affected the result. The best method is to use a variety of different media, several times, measuring regularly, and then calculate any underlying patterns from there. This allows one to check for any outside factors and eliminate them from the experiment, or account for them in the results.
The tests should take place under circumstances that eliminate as many outside factors as possible. If one is testing the speed of light through media, the sensible precautions to take would be to eliminate all but the one source of light being used (so, switch off all other light sources or block them from the testing area) and to ensure that the light beam is being used, that the light impacts the media as early as possible, to prevent potential impurities int he air from affecting the results, and to ensure that there are no obstructions. Anything that could affect the results that isn't the thing being measured should be eliminated.
It is often best to continue an experiement for an extended period of time. Certain outside factors may be difficult to eliminate, and may be dependent on the passage of time, whether it be of a seasonal nature or otherwise. Therefore, tests conducted over years with repeated vigorous result-gathering exercises are more accurate and reliable.
Overall then, for the best results one would;
- Select a large test group
- Eliminate outside variables or account for them
- Take thorough, repeated and regular measurements
- Continue the study for a significant period of time
- Take similar test subjects.
So how do the above matters reflect upon this particular experiment? We know that the group being studied is prone to sudden and large weight, height, hormone and shape fluctuations. We know that there is very little we can do to predict when these will begin or how severe they will be. We know that the goal of the study was to determine whether reading about weight loss or healthy living can result in changes in lifestyle that positively affect BMI.
I would expect this test to consider the following;
- A significant margin of error in all calculations to account for puberty-based bodily fluctuations
- A significant group size to eliminate outlying variables
- An upper and lower starting height, weight and BMI for all test subjects to eliminate the chance that an already significantly larger or smaller test subject could affect the results in either direction
- Regular measurements, at least weekly, to ensure that a standard trend in BMI changes over a long period, but also monthly for bodily cycles, is identified for each test subject. (In that manner, we would only note as significant any BMI changes that lie outside of the cycle already identified, also accounting for the margin of error)
- A significant study period, to eliminate the possibility that BMI fluctuations at any measurement stage may be effected by seasonal changes, puberty, monthly cycles, growth spurts and other factors not predicted
- Evidence that possible outside factors were eliminated from the results.
1; We have been provided no margin of error from which to objectively study the results. However, considering the factors already stated above, one would expect any announced successful "gain" or "loss" of BMI to be of significant amount. This study resulted in maximum increases in BMI of 0.5% and maximum decreases in BMI of 0.71%, with an overall loss in BMI of 0.33% in the control group. This seems extremely small at first glance. More on this later.
2; The total group size was 64 children, divided into 3 groups. Firstly, the groups could not have been of equal size since 64 does not neatly divide by 3. Secondly, this woulds mean approximately 20 children were in each group. If a single child in one group experienced more significant gains or losses than the rest, this could have a massive affect on the results. This is a very small test group.
3; We have been provided no indication of the requirements for the test group, so cannot determine how similar the test subjects were at the start of the test.
4; The groups were only weighed twice during the entire study; once at the beginning, and once at the end. That is disgusting.
5; The test only took place over a period of 6 months. Considering the massive bodily changes many of the test subjects may have undergone during this period, this seems a woefully inadequate period.
6; The group all had 2 things in common. Firstly, that they were already considere obese and secondly, they were already enrolled in another weight loss programme.
Regarding point 1; we have very unhelpfully not been advised the actual figures as measured. We don't now what the starting heights, weights and BMI of the test subjects were, so we don't know how much the actual changes represent. However, we can make a calculation. I regret that I have been unable to search out any figures for "obese" children in terms of weight and height charts, but this page has a chart for the averages, which I will use for the purpose of calculating some figures.
Let us take an 11 year old girl, being 56" in height and approximately 81lb in weight. In other words, 144cm height and 36kg weight. If this child were to increase their BMI, over a period of 6 months, we can generate the following potential height/weight changes.
At the extreme ends of the study, discounting the utterly unknown margin of error, the increase in BMI of 0.5% could represent an increase of 0.180kg in weight, or a loss in 0.004metres in height (0.4mm). In other words, this increase could be the result of a very tiny weight increase or a statistical error in height measurement, depending on the number of decimals to which height was measured (or a child wearing thinner socks the second time around).
The decrease in BMI of 0.71% could represent a decrease in 0.255kg in weight, or an increase in 0.005m in height (0.5mm).
Considering that this study was taking measurements of children going through puberty, I would expect that these actual differences would be wildly smaller than the margin of error and therefore UTTERLY NEGLIGIBLE.