Re: Re:
FrankDay said:
LOL. Welcome back King, I thought you might have dropped off the end of the earth since you didn't respond to my challenge to come here and educate us as to how to properly interpret a p-value given a real world example. Of course, I knew that to not be the case since you were posting on other threads. I interpreted your failure simply to represent you didn't want to post "the truth" because it effectively back what I am saying. But, alas, when you do return, all you do is point out that a small aspect of what I said is wrong without saying what would be more proper to say or to address the original question to you. Anyhow, here is another chance.
No it is not a small aspect, it is the basis of your whole argument that is wrong. I didn't respond to your question because it has absolutely nothing to do with what I was telling you, that p-values cannot be used to assign a probability.
So, given the two examples I used how should the lay person reading this thread interpret those to data results. In one instance two groups are compared and the difference in power and efficiency change between the two groups reached the 0.125 and 0.25 level. What should that mean to the reader?
There is no such thing as the 0.125 or 0.25 "level". A p-value for significance is chosen BEFORE any stats are done, not after. I'm assuming the authors chose 0.05? The only STATISTICAL interpretation of those number is that they are not statistically different assuming a 0.05 level, if it was chosen BEFORE analysis. This is just another example of how you either have no idea about the statistical interpretation or are purposefully attempting to cherry-pick things you wrongly think enforce your argument due to your complete misunderstanding of the statistics.
I do not care how you interpret that data. You can say you think they are the most brilliant data in the world. what you cannot do is try and use a p-value to assign a percentage of likely significance. You can discuss the design, the real numbers, why you think they are meaningful but you cannot falsely attempt to back it up with a misinterpretation of the statistics.
In general terms you seriously expect someone to assess a whole experiment based on two numbers without even giving the relevant constraints? That is ridiculous. They can ONLY assess it based on whatever level was chosen and cannot offer any other insight into the data.
Then, in the crank length study what does it mean (in real world terms) that the 145 and 170 crank lengths reached the p<0.05 level compared to 120 and 195 cranks? And, in view of there being an obvious difference between 145 and 170 mm cranks there must be a p value comparing those differences. If Martin had given us that value how should that be interpreted as regards any difference in crank length in power generation as determined by this study?
Firstly, it is 120 and 220, not 120 and 195. It's right there in the graph you posted.
Secondly, I don't care, again this has nothing to do with our previous discussion and I would hazard this thread at all. How was the p-value calculated? What was compared to what? ANOVA? T-test? Were they corrected? If it's a t-test your whole question is wrong as it cannot compare two populations against two other populations. Were the populations combined? How does this relate to the initial hypothesis? Why must there be a p-value comparing 145 and 170? Was it calculated? Why are they obviously different? The study appears to only care about power from what little you presented, so they can only be assessed in such terms. If the power is not statistically different, in terms of this study, statistically, the cranks are not different either. Just like if I painted a set of cranks red. They would obviously be different to black ones, but not in terms of this experiment if power output is the same. Even if you can reject the null hypothesis, this does not mean that the alternative hypothesis is true because the statistics have no idea what the hypotheses are, only those educated in the field can decide that. The very people who are disagreeing with you (Not me. I am an analytical biochemist and I have only been pointing out your misappropriation of p-values).
Don't answer those questions, I've already pointed out I don't care and it has nothing to do with our previous discussion. I'm just pointing out how silly it is to expect anyone to interpret a whole experiment based on one number (not even a number) and a supposition.
By the way, I agree that the p doesn't represent the probability that the null hypothesis is true. In fact, it is unlikely that the null hypothesis is ever true since the null hypothesis is the two groups are the same and this would never be the same. For the null hypothesis to be the same both groups would have to have the exact same result and the exact same standard deviation. I guess that is possible but the likelyhood is tiny, tiny, tiny.
No, this is again wrong. The null hypothesis is that there is no statistical difference between the groups, not that they are exactly the same, it is the result is due to chance alone. The observed populations can vary and are expected to, there would be absolutely no need to calculate a p-value if your hypothesis were these are different, you'd just look at the numbers. The whole point of a p-value is to give some rigor in how those numbers are interpreted BECAUSE of that variation.
However, that is how it is generally interpreted as these studies are designed to determine if a certain thing has an affect. If the differences are small it is reasonable to assume the groups are the same. But, you are right, it is not technically correct. Good to point out in your statistics class but perhaps a bit over the top for this forum. That is why we are relying on you to tell those of us who don't have your background how to properly interpret and talk about this kind of data.
It is not a small point, and it is not how they are interpreted. It is completely fundamental to any interpretation of a p-value and seems to be the basis of your misunderstanding. The populations can vary wildly, the null hypothesis is that any variation is due to chance alone.
You're not relying on me at all. If you were then this conversation would have lasted for about three posts. Your initial post, mine telling you that you cannot interpret a p-value in that way and possibly your response saying thanks (I say possibly because I wouldn't have expected it, this is not a personal dig). That would have been it. Instead you have constantly mis-interpreted references, at times falsely edited them to fit your argument and tried to engage me in discussions I do not want to have and that have no relevance to the point being discussed.
My ONLY interest in this whole conversation is that you stop falsely using p-values. Other than that I do not care how you interpret the data, I am no way qualified to assess that.
The people you should be relying on for that are the coaches and exercise physiologists. Only they have the relevant experience and knowledge to interpret the data beyond the statistics and explain why it is happening. But, unfortunately, they don't agree with you.