View Post

[Torillian]

DarkD said:

Torillian said:

I'm not a social scientist so anyone with more experience in this field please correct me, but my understanding of the study in question is that the null hypothesis in this case was that socially youth and nontransgender youth have similar levels of anxiety. In this case a higher p value means that the two groups did not show significant differences in their depression which is the conclusion being made. 

A quote from the study that this one was following up on:

"In terms of depression, transgender children’s symptoms (M = 50.1) did not differ from the population average, P = .883. In contrast, transgender children had elevated rates of anxiety compared with the population average (M = 54.2), t(72) = 4.05, P < .001"

So the children did not have differing levels of depression, but had statistically significant differences in their levels of anxiety. Basically, this study having high p-values is the reason they conclude that socially accepted transgender youth do not have differing levels of depression from cisgender youth. To throw out the study because of high p-values is to misunderstand the study itself. 

The P-value means the margin for error in the conclusions drawn.  So a .05 means it's acceptable as scientific evidence.  Anything above that makes it unacceptable.  You can still draw conclusions from it, but it will not be taken as scientific evidence unless the p-value is lower than .05.  

The study has such a bad p-value because the sample size is so small.  Only like 50 patients.  Even a sample of 1000 patients would be iffy by these standards and may not pass the p-value test.  So yea, the study doesn't mean anything.  

In fact, the conclusions with the best p-value are the conclusions that favor the conservative point of view.  "Trans reported marginally higher anxiety compared to the control group p=0.076" and "trans reported marginally higher anxiety than the national average =0.096" and finally, the only scientifically relevant fact present "parents reported their children had more anxiety than children in the control groups =0.002"

So the only facts this study brought to the table are that trans children have anxiety problems.  

Alright I'll try this once more. The P-value in these studies is an attempt to show a statistically significant difference between the groups investigated. Larger p-values mean there are not statistically significant differences for that value. There is no way that the statement "these two groups have similar depression levels" could be proven by a low p-value because a low p-value would show a difference between the two groups. The fact that the two groups had high p-values for the measured value of depression is the reason they conclude the two groups have the same level of depression. If the p-value had been low then they would conclude that the two groups are different. There is no way to conclude the two groups are the same with a low p-value, it's just not how the analysis works.

That is my reading on how the studies are designed, if you have more information on this than I I'd be interested in hearing it, but simply repeating the idea that only conclusions with low p-values are valid is meaningless. p-value is not "the error in the conclusions drawn", but a means of showing whether a null hypothesis has been disproven. If the null hypothesis is that the two groups are the same, and your p-value is high this means that the two groups are not different to a statistically significant degree. This is what the study stated. 

So, given my reading of the study, what are your credentials to be discrediting the analysis of experts in their field? Because I'm pretty sure these guys understand P-values much better than you or I. 

Here's some info to back up my interpretation:

https://www.biochemia-medica.com/assets/images/upload/xml_tif/Marusteri_M_-_Comparing_groups_for_statistical_differences.pdf

Last edited by Torillian - on 26 August 2019