Correlation, Predication, Confidence, and Errors: Analysis of Statistics in the News

In statistics, there are a number of ways to compare how data is associated with the subject matters being discussed. In this paper, I will discuss how three different magazine articles use statistics to improve and expound on their theories, the methodology used, and if their use of statistics made the article more or less convincing. The statistical methods that will discuss statistics in the news include correlation, predication, confidence, and errors. Correlation “exists between two variables when higher values of one variable consistently go with higher values of another variable or when higher values of one variable consistently go with lower values of another variable. (Bennet, Briggs & Triola 2014)” It is only used with quantifiable data where the numbers are meaningful (height, weight, age, etc), therefore, it can’t be used for categorical data (gender, favorite foods, jobs, etc). Predication is defined by Britannica.com as the attribution of characteristics to a subject to produce something meaningful. This combines verbal elements and those that exist in name only (Britannica 2015). Confidence in statistics is defined as “a group of continuous or discrete adjacent values used to estimate a statistical parameter as a means of variance” (Merriam-Webster 2015). Statistical errors means that when null hypothesis is proven incorrect, then the alternative hypothesis is accepted (Rogers 2015). Errors are broken into two types, based on the null hypothesis where the assumption is that the treatment didn’t do anything at all. It could be called the base zero where nothing has been proven or disproven. A Type I error detects an effect that isn’t present and Type II fails to detect a present effect. Also sometimes called false positives and false negatives. (UTexas.edu 2011)

Why having kids is bad for your health (Rochman 2011)
The article in Time magazine titled Why Having Kids is Bad for Your Health argues that parenting “may be hazardous to your health”. It supports that statement with new research from the University of Minnesota which looked at 838 women and 682 men that show negative results occur when one has children, especially for the mothers. The article states that mothers had a higher body mass index and ate a less healthy died with more sugary drinks and higher calorie and fat foods than women without children. (Rochman 2011) But the article also states that a corresponding BMI in fathers was unaffected, but didn’t give a theory for why that might have occurred.
The lead researcher who conducted the study, Jerica Berge, stated that the difference between mothers who ate more calories and childless women who ate fewer calories was “significant” (Rochman 2011), although when asked why it was significant, Berge wasn’t able, or willing, to establish the correlation or significance but didn’t disagree with the reporter when it was suggested that mothers might be eating the children’s leftovers which would be a potential causal relationship.
The text from Statistical Reasoning suggests that a correlation exists when higher values of one variable exist with higher values of another. However the article clearly states that although mothers and fathers were both parents to the children, the only one with significant negative outcomes were mothers without a corresponding outcome for fathers. And when pressed, the researcher would not speculate on what made the outcome significant.
The author, who is herself a mother of two small children, mentions the relentless pressures of parenting and lack of time or energy to cook more healthy meals as a reason for the higher intake of calories and lack of daily exercise as compared to childless women. While this does sound like a plausible argument, it still doesn’t take into account the difference between mothers and fathers and no other identifiers were mentioned in the article as to the differences between the parent’s diets and exercise routines that could also be causing the differences. I think this would be a good example of a Type I statistical error.
The article makes a statement about mothers and offers a couple of different theories about why this could have taken place, but the theories were not mentioned in the study so were not part of the testing procedure so a true correlation cannot take place. The article also mentioned mothers and non-mothers, as well as mothers and fathers, but didn’t give any statistical data to infer a causal or corresponding relationship. The initial research was stated as being conducted on women and men, but did not mention how many of the women and men tested were parents. So the conclusions drawn between mothers and non-mothers, and between mothers and fathers, is not established by the facts portrayed in the article. Therefore the article’s argument that having kids is bad for your health is not convincing since additional evidence on the effect on fathers would need to take place, and an explanation from the researcher explaining what makes the difference significant would need to be addressed and possibly tested. If the relationship is causal and corresponding, more specific statistical evidence would need to be in place and reported in the article.
While the article does offer some seemingly good advice on ways to get more physical activity into a mother’s daily life, or to help with time management that could affect the time used to cook healthier food, I would not use this article as a reason to make changes in my life because a causal relationship hasn’t been established and much of the theory seems to be anecdotal on behalf of the article’s writer rather than taken from the actual analysis this is purported to be based on. It seems like a very flimsy argument based on a reporter’s assumptions of the report than from evidence in the report itself.

Why going to church can make you fat (Park 2011)
The second article, also from Time, called “Why going to church can make you fat” (Park 2011) attempts to describe a study that found that people who attended religious activities gain more weight then those who don’t go to church as often. The evidence used in the article to establish the correlation between attendance at religious events and weight gain over time was a study by researchers at Northwestern University. The researchers analyzed data from a study that followed more than 2,400 people aged 20 to 32, for a period of 18 years. That study found that people who attended church at least once a week were more than twice as likely to be obese as those with no religious activity. (Park 2011) The authors speculated that the rise in obesity could be a result of more church-based social activity that could potentially center around food.
While there could be a correlation between the outcome of weight gain and attendance at religious events, especially given the survey parameters and length of time devoted to the study, there were many factors that aren’t mentioned such as family history, gender, amount of exercise, lifestyle, geographical area, etc. that could also have led to the weight gain other than simply attending weekly religious services. According to the article, the researchers speculated that the weight gain was due to broader and more active social networks, but did not address which activities were associated with the most weight gain (Park 2011). The article also mentioned several potential ways to use the church social atmosphere to combat the weight gain.
To establish a true correlation, researchers would need to address what social activities the church goers were attending to see if there was evidence that those with the most weight gain actually attended the social opportunities. There would also need to be a more intense study of other possible contributing factors. While the evidence is compelling, I do not believe it is convincing due to it’s very broad scope and limited reported statistical evidence. The article also mentions several ways the religious groups could benefit from the same social networks to lose weight by using the groups as a support system for diet and exercise.
If the researchers wanted to convince, they could provide more evidence from the in-depth data they analyzed from the Coronary Artery Risk Development in Young Adults study. They could also offer supporting evidence that group diet and exercise programs within the framework of a religious organization would be successful. There is no evidence within the article that says whether that theory was ever tested or even mentioned to the survey participants, or would be a factor in either weight loss or lifestyle changes. The article also mentions that the researchers have been working with a church to address these issues, but gave no indication of whether or not their perceived solutions provided measurable results.
The article mentions other unnamed studies that would support church-based health initiatives that seem to show that church-goers already have better overall mental and physical health than non-churchgoers, but admit that obesity is an area that still has room for improvement. While the assumptions pointing to using church-based activities to propel weight gain, exercise, and fitness routines makes a lot of sense, the evidence of a causal relationship between religious attendance and weight gain is flimsy at best, and needs a great deal more research to find the correlation of what within a weekly church attendance would be the actual contributing factors in the attendees weight gain.

The link between sleep and weight (McCoy 2010)
This article seeks to address the question of if getting enough sleep at night can help an individual gain weight. The evidence provided for this article was based on three recent studies: the National Sleep Foundation survey, another unnamed study that tracked a group of 40-60-year-old women for five to seven years and tracked their weight and sleeping patterns, and another unnamed study that examined the eating and exercise habits of a group of healthy young men over two nights. Three additional studies found a correlation between less sleep and childhood obesity.
The argument that the relationship between sleep and diet is causal is that “people who are sleep-deprived may have less energy throughout the day and therefore less motivation to exercise.” (McCoy 2011) In addition, “the amount of sleep you get seems to affect the appetite-controlling hormones ghreling and leptin, which can leave you reaching for high-carb and calorie-dense foods when you haven’t gotten enough sleep.” (McCoy 2011). Sleep deprivation leads to higher levels of ghrelin, which stimulates appetite, and lower levels of leptin which helps suppress the appetite.
This article seems to be well researched and shows the cause-and-effect relationship between sleep and diet very well, which gives confidence in the overall conclusions. While several of the studies she mentions are not named, a couple of the specific studies were conducted by well-known, and highly regarded institutes which helps give a level of credibility to the results that were missing in the previous two articles. The author also doesn’t attempt to introduce new information that isn’t supported by the research at hand. Her article is well research and documented, and chunked in a seamless way that makes it both readable and helps increase understanding of the subject matter. She also mentions that the research is compelling that an individual should get enough sleep to maintain a healthy weight.
The sources of her research and the measured way she relates the research to her argument was very interesting and seems to support the correlation between sleep and weight loss. I find this article to be credible and since both sleep deprivation and weight gain are issues that I struggle with in my own life, this will compel me to ensure I get enough sleep at night in order to see if it helps with weight loss in the future.

Conclusion
Research and statistics are used by the media to explain and support claims they make in their articles and papers. Some use these statistics in ways that do not support or correlate with the argument they are trying to make or use assumptions and guesses based on erroneous or misunderstood information to distort the findings of scientific research. Good journalism uses the statistics in ethical and meaningful ways, allowing the full facts to lead the reader to a better understanding of the issues. Bad journalism takes striking statistics and makes leaps to give conclusions and “facts” that are not supported by the evidence. These unsupported leaps in correlation and causation can decrease confidence in the writing and render its conclusions irrelevant.

References:

Creative Research Systems (2015) The survey system: correlation. Retrieved from http://www.surveysystem.com/correlation.htm
Bennett, J., Briggs, W. L., and Triola, M.F. Statistical Reasoning: For Everyday Life, 4th Edition. Pearson, 2014. VitalBook file.
Encyclopedia Britannica (2015) Predication: logic. Retrieved from http://www.britannica.com/topic/predication
Merriam-Webster (2015) Confidence interval. Retrieved from http://www.merriam-webster.com/dictionary/confidence%20interval
Park, A. (24 March 2011) Why going to church can make you fat. Time Magazine. Retrieved from http://healthland.time.com/2011/03/24/why-going-to-church-can-make-you-fat/
Rochman, B. (11 April, 2011) Why having kids is bad for your health. Time Magazine. Retrieved from http://healthland.time.com/2011/04/11/is-parenthood-bad-for-your-health/?iid=WBeditorspicks
Rogers, T. (2015) Type I and Type II errors – making mistakes in the justice system. Retrieved from http://www.intuitor.com/statistics/T1T2Errors.html
UTexas.edu (12 May 2011) Type I and Type II errors and significance levels. Retrieved from http://www.ma.utexas.edu/users/mks/statmistakes/errortypes.html