Perception bias is pervasive throughout your link regarding Cuomo's culpability. If intended as a defensive against his negligence, it fails miserably in denying his sending thousands of "typhoid Marys" into ALFs.
As to unbiased and objective news sources, that you seemingly singled out for criticism a few while clearly omitting others displays your bias too.
I don’t mean to imply bias is a bad thing; only that it needs to be recognized in all of us. It’s how we deal with it that is important.
Let me begin by saying I'm neither attacking or defending Cuomo. The article I cited is not particularly well written, however, it does present opposing views and offers appropriate critiques. For example this quote from the
Post-Standard Article:
“It’s important to be cautious about reading too much into it,’' said Bill Hammond, senior fellow for health policy at the Empire Center. “First of all, correlation does not equal causation. There are confounding factors that could be clouding the situation.’'
It is worth noting and the P-S article notes the Empire Center is a conservative policy center in NY. It is accurately stating the limitation of this study. This is commendable. The P-S article goes on to identify other factors that could have an equal or greater effect on nursing home deaths that may be clouding the data. Again, this is commendable as the author is trying to present a balanced and fair report.
I am either blessed or cursed with more knowledge of advanced statistics than the average person, having taken and passed some 18 graduate credits in inferential statices, including multiple regressions, which the
Empire Center's report used. With this background, I have a different understanding of the data, and frankly the data is pretty unconvincing for several reasons.
The data is correlational, i.e., it shows a correlation between the policy and returning covid patients to SNF. While tempting to believe that one caused the other, any one who took an entry level course in statistical analysis learns correlation does not imply causation, it simply demonstrates that both variables (deaths and readmits) vary in the same direction. There may or may not be a relationship. A good example of this type of data is touted in late January every year
when forecasters remind us of the relationship between the Stock Market and who wins the Super Bowl.
The Empire Center data shows a "statistically signifiant" relationship between readmits and deaths. This sounds more impressive than it is for a couple of reasons. First, statistically significant relationship only states the probability of obtaining a correlation given the data and the size of the data set. Notably, the EC Report does not indicate at what level significance was established, it could be as high as .1% or as low as 15% That is the range typically used by researchers, with very high levels of sigificance (.1%, ) for important and risky decisions, such as in drug studies or vaccine studies to very low levels fo significance (15%) for less critical or exploratory studies, sometimes known as fishing expeditions.
When conducting a correlational study the larger the group that is studied, the easier it is to find statistical relationships and significance is option with a much lower correlation. If the data set reaches hundreds and thousands of data points, the correlation value can be quite low and still reach statistical significance.
Mathematically the correlation is the square root of the variance. So what does that mean? In correlational studies we often talk about shared variance, or what unknown factors are involved in arriving at the correlation. If you recall Venn diagrams, there are 2 or more overlapping circles, the area of overlap is the shared variance. To give an example, a popular type of study and part of the validation procedure for achievement tests and IQ tests is a correlational study to see how they relate. Typically these studies yield a correlation of about
.y, .7 which is a pretty strong correlation, it says the better you do in an IQ test, the better you are likely to on a reading test. Makes sense, but if we look at the shared variance it is at best about .49 or 49%. That means at least half of the factors that contribute to good reading skills are the same that contribute to high IQ. It also means that half of the reason has nothing to do with IQ.
The EC Report states:
The data show that each new admission of a COVID-positive patient correlated with .09 additional deaths, with a margin of error (MOE) of plus or minus 0.05.
First, all correlation values fall between 1 and -1, with a value of 0 meaning there is no correlation. A value of .09 is very close to 0 which suggests there is very little relationship between between deaths and readmits. The shared variance is .0081 or less than 1% of what ever caused the deaths. A truly trivial amount.
To go back to my original point, media with an agenda, like Fox, ONAN, Sinclair Broadcasting, all the Right Wing Talk show hosts, will never go into depth about or accurately report findings like this. At least the P-S article tried to report it, but the level of statistical sophistication is beyond that of most reporters.
Finally, the media with a known bias will never report these paragraphs from the EC Report:
As with any such analysis, the results should be viewed with caution. Even a statistically significant correlation between two factors does not necessarily mean that one caused the other. The available data were also limited in potentially important ways—such as the lack of dates for the COVID-positive admissions.
Other possibly relevant factors, such as the relative quality of care provided in the nursing homes and the average acuity of their patients’ condition, were beyond the scope of this review. Moreover, the data do not clarify how many of the patients admitted to a nursing home from a hospital later died in the nursing home, which would add to the home’s death count even if the patient in question did not spread the virus there. (emphasis added)
Edit: See strike through above. Just a typo, .y is a meaningless term, it should have been .7. Proofreading is not one of my strong suits.