
Welcome to the International Skeptics Forum, where we discuss skepticism, critical thinking, the paranormal and science in a friendly but lively way. You are currently viewing the forum as a guest, which means you are missing out on discussing matters that are of interest to you. Please consider registering so you can gain full use of the forum features and interact with other Members. Registration is simple, fast and free! Click here to register today. 
19th April 2017, 05:51 AM  #241 
Pi
Join Date: Nov 2005
Posts: 14,792

Is an EM drive being tested aboard an X37B?
I'm guessing not. https://www.sciencealert.com/rumours...justwontdie 
__________________
Up the River! 

19th April 2017, 11:46 AM  #242 
Graduate Poster
Join Date: Sep 2004
Location: USA
Posts: 1,354


21st April 2017, 12:16 AM  #243 
Penultimate Amazing
Join Date: Jun 2009
Posts: 10,489


__________________
لا إله إلا رجل والعلوم والتكنولوجيا وأنبيائه 

23rd April 2017, 06:29 PM  #244 
Philosopher
Join Date: Jun 2010
Posts: 6,042


24th April 2017, 03:27 AM  #245 
Penultimate Amazing
Join Date: Jun 2009
Posts: 10,489

Hm, it says the thrust at 40W is 30 ± 6μN and rises to 106 ± 6 μN at 60W, but drops back to 76 ± 6μN at 80W.
As stated there would be more thrust at 60W than at 40W. The fact there is less thrust at 80W than at 60W leads me to believe their confidence intervals are bollocks, but the authors stated there was a measurable and statistically meaningful increase in thrust. Or am I missing something? McHrozni 
__________________
لا إله إلا رجل والعلوم والتكنولوجيا وأنبيائه 

24th April 2017, 04:19 AM  #246 
Philosopher
Join Date: Jun 2010
Posts: 6,042

The instrument error is 6 uN*. The measurements produced by that instrument are all over the place. That is what gets factored into confidence intervals. Those should be the purple whiskers behind each grouping, although those don't appear to be confidence intervals either  a quick back of the envelope says the SEM of the 60W group should be + 23 uN, not the fiftysomething they label.
*Almost certainly determined using optimal conditions, and not something on the bare edge of detection. 
24th April 2017, 04:34 AM  #247 
Penultimate Amazing
Join Date: Jun 2009
Posts: 10,489

Yeah OK, that's a bit different.
Still, 106 ± 23μN is larger than 30 ± 23μN. The true uncertainty would have to be about ± 38μN for the two results to be statistically indistinguishable. My take is that it is indeed either indeed this large or that the large measurement at 60W is bogus in some other way. McHrozni 
__________________
لا إله إلا رجل والعلوم والتكنولوجيا وأنبيائه 

24th April 2017, 10:17 AM  #248 
Philosopher
Join Date: Jun 2010
Posts: 6,042

That's still not quite right. 23 uN is the confidence interval around the average of their 60W measurements, not something applied to each of the measurements or the other wattages.
To use an analogy, have you ever noticed that when you take your temperature multiple times in a row, you'll get slightly different numbers each time? That happens because of a lot of factors  instrument error, how you're holding it, where you're holding it, etc, but each measurement is helping create a sample of your unknown true temperature. Do it a couple of times and you can start averaging the measurements together and using their spread to predict how accurate the mean probably is, mathematically known as the Standard Error of the Mean. A little less than double that will get you a 95% confidence window in which your true temperature most likely falls. I worked up the 40W measurements here. 
24th April 2017, 10:53 PM  #249 
Penultimate Amazing
Join Date: Jun 2009
Posts: 10,489

I was under the impression that 106μN and 30μN were average measurements at 60W and 40W, respectively.
Alas, it is true my knowledge of statistics is limited and I acknowledge it needs to be updated Can you maybe point out a good online tool to study it a bit? It would be useful for my work. Thanks McHrozni 
__________________
لا إله إلا رجل والعلوم والتكنولوجيا وأنبيائه 

25th April 2017, 10:31 AM  #250 
Philosopher
Join Date: Jun 2010
Posts: 6,042

Frankly, the best place to start is wikipedia. Its only problem is it sometimes trips over itself trying to present inscrutable mathematical derivations rather than anything which can be reasonably scrut.
Have a look at their standard error page. It has the formulas I used in the code above. Do you need more or less rigorous math than that? 
25th April 2017, 01:55 PM  #252 
Uncritical "thinker"
Join Date: Jan 2007
Location: UK
Posts: 17,599

Not an online tool, but I like the NIST Sematech ehandbook of statistical techniques. I'm on a phone so posting links is a pain but it's worth a Google.
It pretty much describes my approach to exploratory data analysis, which works very well in my sorts of engineering. Basically plot the data in lots of different ways before guessing what statistical tests to perform. 
__________________
OECD healthcare spending Expenditure on healthcare http://www.oecd.org/els/healthsystems/healthdata.htm link is 2015 data (2013 Data below): UK 8.5% of GDP of which 83.3% is public expenditure  7.1% of GDP is public spending US 16.4% of GDP of which 48.2% is public expenditure  7.9% of GDP is public spending 

25th April 2017, 10:51 PM  #253 
Penultimate Amazing
Join Date: Jun 2009
Posts: 10,489


__________________
لا إله إلا رجل والعلوم والتكنولوجيا وأنبيائه 

25th April 2017, 10:52 PM  #254 
Penultimate Amazing
Join Date: Jun 2009
Posts: 10,489


__________________
لا إله إلا رجل والعلوم والتكنولوجيا وأنبيائه 

Bookmarks 
Thread Tools  

