Today at 05:26 PM Eddie said this in Post #24
One at a time. Byrd constructed a well designed study on the impact of incremental prayer against an uncontrolled background. Kudos to him.
Unfortunately, there really is nothing there (I think, the stats he used were a bit unclear).
When you look at the data, you find that Byrd is indeed using a shotgun. That is perfectly fine considering that he is collecting the first data on this subject.
Suffice it to say that running repeated significance tests on an inventory of attributes proves nothing since due to the nature of the underlying distribution one would expect certain number of significant results.
Yes, one would. However, of 26 categories and a significance level of p< 0.05, you would expect 1 or 2 to be significant. Byrd saw 5.
I will see what I can do about reanalyzing the data. I have also sent the data to an accomplished analyst for a look see, but I am virtually certain that I am correct on this.
The data has already been analyzed by Cochrane Review, and they find it sound. Cochrane Reviews take all the papers on a subject and analyze them so that they can summarize the results for physicians and present them with recommendations. I'll be happy to e-mail you the review if you would like.
A point to remember is that doctors tend not to be researchers.
I know that all too well.
After all, I'm a Ph.D. in a clinical department. But in this case Byrd had already had 4 papers published in the peer-reviewed literature. You can find them, like I did, on PubMed. Byrd simply applied the same methodology he had used in drug testing on intercessory prayer, using prayer as a drug.
My analysis of the Results is below:
RESULTS
Statistical analysis was done on the medical condition of the patients at time of admission. There was no statistically significant difference between the medical conditions of patients in the control group and the prayer group. This is important. The 2 groups started out the same. The parameters measured included age, gender, primary cardiac diagnosis, and preliminary noncardiac medical conditions (diabetes, pneumonia, drug overdose, etc.) A total of 30 medical parameters were studied at entry. As far as I can tell they cover the relevant areas. (I would think the reviewers would have picked this up if it were a problem.)
There was no difference in days in CCU, days in hospital, or number of drugs that needed to be taken after discharge between the two groups. While the patients were in the hospital, 26 categories of new problems, diagnoses, and therapeutic effects were measured. These included mortality (the patient died), congestive heart failure, need to be put on various drugs, hypotension (low blood pressure), major surgery, arrhythmia (irregular heart beats), etc. Again, the categories seem relevant.
Of the 26 categories, congestive heart failure, cardiopulmonary arrest (new heart attack), pneumonia, diuretics (drugs to control high blood pressure), antibiotics, and intubation/ventilation were seen less frequently in the prayer group at the statistically significant level of p < 0.05. This is the level considered statistically significant in all other scientific studies. What it means is that if you took all the patients and picked patients at random into 2 groups, 1 time out of 20 you would get 2 groups that would give these results.
The significance of the 6 parameters that were statistically different does not tell the whole story. In 14 other parameters, the prayer group was less than the controls, in one they were the same, and were worse in only 5 areas. Also, the chi-square test is not that sensitive. For instance, for diuretics 5 patients in the prayer group needed them compared to 15 in the controls. That is statistically different. But for major surgery before discharge 5 patients in the prayer group and 14 in the control group had it. That difference of 1 patient means it was not quite statistically different. Or in congestive heart failure 8 patients in the prayer group vs 20 in the control was statistically significant. In the category of receiving ionotropic agents (drugs to help the heart beat) 8 patients in the prayer group compared to 16 in the control was not statistically different. For antianginal agents 21 patients in the prayer group needed them compared to 19 in the controls.
Thus there is a trend. Where the prayer group is better, in 14 categories it is a lot better by 30-50% fewer patients. In the 5 categories where prayer is "worse" it is only by 1 or 2 patients. Example, 2 patients in the prayer group needed a permanent pacemaker compared to 1 in the control.
Byrd did an multivariant analysis between the groups. This compensates for variations among the areas. For instance, prayer might be significant in antibiotics but controls might be close to significant in a lot of other areas and the overall effect might no difference between the groups. Multivariant analysis will pick that up. In the multivariant analysis the prayer group was significantly different at the p < 0.0001 level. That means that only 1 in ten thousand tries would these results show up by chance. Most of us dream of p values like this. (In a statistically rigorous world, p values should never be compared. You pick your p value at the beginning of the study and then only say results were significant or not significant. However, I have never yet seen a statistically rigorous MD unless there were a statistician standing next to him pointing a gun at his head.)
Byrd also graded the hospital stay as "good", "intermediate", or "bad" based on what happened during the stay. "Good" was when no new problems arose or only those problems that minimally increased the risk of mortality or morbidity, "intermediate" was higher levels of risk, "bad" was the highest morbidity or those who died during the study. 85% of the patients in the prayer group were "good" compared to 73% of the controls, 1% prayer vs 5% control for intermediate, and 14% prayer vs 22% control for bad. Prayer was statistically different from control at p < 0.01.
Now, it appears that Byrd had broken the code when he did this analysis, but it turns out that it is "robust". That is, if you put all the intermediate in the "good" category, or all the intermediates in the "bad" category, the statistical results are the same. Thus, Byrd appears to be 1) an honest grader and 2) even if he shaded a bit, the categorization is so marked that any possible shading doesn't matter.
[/B]
[/QUOTE]