We now have 203 participants. I have not updated the statistics yet, but did a rewrite or the Medical paper (based on feedback and more research on my part) and more personalized one that I am trying to get published in a magazine. Both of these have been added to my web site along with a video of my leg and my BFS. Thank you to Mitra and my wife for their help editing some of these writings.
http://patrickbohan.elementfx.com/BFS.htmAs far as the medical paper is concerned some of my assumptions about statistical significance may not be true, so I will have to add a Spearman correlation model to my results. What was even more troubling was that I calculated what type of sample size we need for our survey to have 5% error and 95% confidence level - 384 is the magic number. We will get there as people join our social network, but it will take time. I may not update any models until we get to this point, since it is a lot work considering I have to add the Spearman correlation models as well. I should have calculated all of this earlier and done a better job researching statistical significance. I apologize for this. For those who do not want to read the whole piece on my web site below is what I added to the medical paper (Statistics can be very confusing):
Sample Size
What is the correct sample size for this survey study? First, we need to determine (estimate) how many people suffer from severe and chronic BFS symptoms (Population Size). Symptoms must be bad enough to for a patient to see a neurologist to be officially diagnosed with BFS after possibly having an EMG and or brain MRI performed to rule out ALS and MS. According to the Center for Disease Control about 1 in 10,000 people in the U.S. have ALS and about 1 in 600 people suffer from Parkinsonâ€™s disease. At these rates, it means as many as 700,000 people around the globe can have ALS and 12 million people can have Parkinsonâ€™s disease. If the rate of chronic BFS is comparable to the rate of ALS and even Parkinsonâ€™s disease, the sample size of the survey would need to be 384 people to tolerate a 5% error and a 95% confidence level. There are dozens of online calculators available to compute and verify these calculations. Our present level of survey participation has approximately a 75% confidence level.
Correlation
Only those variables that show high statistical significance are analyzed for correlation. Since the survey data is based on a rank-order system (ordinal data), the Spearman method of correlation is be used. [9] Once this survey reaches its goal of 384 participants, a Spearman correlation study will be conducted on those parameters that show high statistical significance. Spearman results can be broken down as follows: +/- 0.5 to 1 as strong correlation, +/- 0.3 to 0.5 as moderate correlation, +/- 0.1 to 0.3 as weak correlation, and 0 to +/-0.1 as no correlation. [9]
Discussion
No survey discussion is complete without talking about statistics. Statistics can be vastly confusing and open to many interpretations. Statistical significance between parameters does not imply practical significance. Sometimes commonsense has to be used to determine practical significance from statistical significance. However, in a complicated survey on a complicated subject (in which we know little about) such as BFS this is hard to decipher.
High correlation does not imply statistical significance especially if the sample size is small. Conversely, statistical significance may not imply strong correlation, it may only occur because the sample size is large. Hence, it important to report statistical significance, sample size, and correlation, together as one item, to determine practical significance. [9] In this survey, since the sample size is becoming large, it is probable that the data may indicate strong statistical significance between various parameters, but only weak to moderate correlation â€“ yet this data can be practical significance. Weak correlation, strong statistical significance, and large sample size is usually better than weak statistical significance, small sample size, and strong correlation. [9] My original assumption that if moderate to high statistical significance only yielded weak Spearman correlation, this would be practical significance because BFS is a syndrome we know nothing about. Hence, any correlation weak or strong would be new and helpful information. For this reason, I did not run the Spearman correlation models. After all, if weak correlation was practically significant than why bother with the models. My assumption might be true, but based on [9] the results of Spearman correlation still need to be calculated and added to this writing.
9 DA De Vaus, Analyzing Social Science Data, Sage Publications 2002; p178-85.