- Thank you received: 0
Requiem for Relativity
- Joe Keller
- Offline
- Platinum Member
Less
More
11 years 4 months ago #11112
by Joe Keller
Replied by Joe Keller on topic Reply from
Several errors to correct: the program as posted switches RA & Decl in the printout; that is, the order of these two numbers is switched compared to what the printout says they are. Also, yesterday's post says 1905 were included but really they weren't. I had changed them to REM statements and forgot to unchange them. More serious, though trivial errors: I discovered July 23, was that I forgot to convert degrees to radians in my precession adjustment subroutine. I discovered July 31, that I had entered grossly wrong ordinates for my reference second harmonic sine wave, and also that my detrending procedure was slightly erroneous. So, here are my corrected results as of July 31:
The best fitting (least sum of squared differences of observed and predicted second harmonic coefficients for each set) drift directions (RA, Decl) with correlation coefficient of observed & expected, sigma, apparent drift speed, no. of sets, no. of turns, and mean Julian Date weighted by no. of turns:
1905* (one July set, four Oct-Nov sets)
210, 25
+0.74,+2.52,11.6km/s
5,230,2417119.880
Apr 1922
0,85
+0.68,+2.19,34.8km/s
5,81,2423162.415
Aug-Sep 1923
15,30
+0.44,+2.73,12.6km/s
18,263,2423660.527
Jun-Jul 1924
90,80
+0.50,+4.93,12.3km/s
41,608,2423979.216
Apr 1927
(There was a (-) correlation that was larger, -0.34.)
265,45
+0.26,+1.62,7.2km/s
20,400,2424984.541
Aug 1927
90,80
+0.45,+4.22,10.3km/s
40,800,2425110.636
Sep-Oct 1929
220,65
+0.67,+3.51,13.0km/s
11,220,2425887.935
All data, 1905-1929 for the steel interferometer at Cleveland:
196,82
+0.2392,+4.060,9.945km/s
140,2602,2423979.047
* For the data subsets, I made only a coarse search on a 5x5deg northern hemisphere (including equator) grid with rough sec(Decl) correction in RA steps. For the whole data set, this was followed by strict 1x1deg search on the largest grid square centered on the best coarse point and not overlapping other coarse points.
I had reason to doubt the sign of the 1905 data (ambiguous notation in Miller's notes, giving me two possible interpretations) so I chose the interpretation for 1905 giving the largest positive correlation. As noted above, I could have obtained a larger positive correlation by reversing the sign of the Apr 1927 data but lacking any other reason, did not do so. Likewise although a convenient way to explain the three times greater apparent drift speed in 1922 would be to assume that Miller was recording hundredths, not tenths, of a fringe that year, I did not make this emendation either.
The coarse search on the whole data set found sigma = +4.032 for RA,Decl = 210,80. So, the significance, p, was almost as great, for a direction three degrees away from the best, suggesting by extrapolation of sigma and integration weighted by 1/p, an error bar of ~15deg.
The Hicks first harmonic correlation was much larger when I assumed that the Hicks effect should be 90deg out of phase with the drift azimuth. If the Hicks effect leads by 90deg, then the largest correlation coefficient is +0.15, sigma = +2.55, at RA,Decl = 339,71; this is 26 deg from the best second harmonic indicated direction.
The best fitting (least sum of squared differences of observed and predicted second harmonic coefficients for each set) drift directions (RA, Decl) with correlation coefficient of observed & expected, sigma, apparent drift speed, no. of sets, no. of turns, and mean Julian Date weighted by no. of turns:
1905* (one July set, four Oct-Nov sets)
210, 25
+0.74,+2.52,11.6km/s
5,230,2417119.880
Apr 1922
0,85
+0.68,+2.19,34.8km/s
5,81,2423162.415
Aug-Sep 1923
15,30
+0.44,+2.73,12.6km/s
18,263,2423660.527
Jun-Jul 1924
90,80
+0.50,+4.93,12.3km/s
41,608,2423979.216
Apr 1927
(There was a (-) correlation that was larger, -0.34.)
265,45
+0.26,+1.62,7.2km/s
20,400,2424984.541
Aug 1927
90,80
+0.45,+4.22,10.3km/s
40,800,2425110.636
Sep-Oct 1929
220,65
+0.67,+3.51,13.0km/s
11,220,2425887.935
All data, 1905-1929 for the steel interferometer at Cleveland:
196,82
+0.2392,+4.060,9.945km/s
140,2602,2423979.047
* For the data subsets, I made only a coarse search on a 5x5deg northern hemisphere (including equator) grid with rough sec(Decl) correction in RA steps. For the whole data set, this was followed by strict 1x1deg search on the largest grid square centered on the best coarse point and not overlapping other coarse points.
I had reason to doubt the sign of the 1905 data (ambiguous notation in Miller's notes, giving me two possible interpretations) so I chose the interpretation for 1905 giving the largest positive correlation. As noted above, I could have obtained a larger positive correlation by reversing the sign of the Apr 1927 data but lacking any other reason, did not do so. Likewise although a convenient way to explain the three times greater apparent drift speed in 1922 would be to assume that Miller was recording hundredths, not tenths, of a fringe that year, I did not make this emendation either.
The coarse search on the whole data set found sigma = +4.032 for RA,Decl = 210,80. So, the significance, p, was almost as great, for a direction three degrees away from the best, suggesting by extrapolation of sigma and integration weighted by 1/p, an error bar of ~15deg.
The Hicks first harmonic correlation was much larger when I assumed that the Hicks effect should be 90deg out of phase with the drift azimuth. If the Hicks effect leads by 90deg, then the largest correlation coefficient is +0.15, sigma = +2.55, at RA,Decl = 339,71; this is 26 deg from the best second harmonic indicated direction.
Please Log in or Create an account to join the conversation.
- Joe Keller
- Offline
- Platinum Member
Less
More
- Thank you received: 0
11 years 3 months ago #13967
by Joe Keller
Replied by Joe Keller on topic Reply from
Regarding the difference between my analysis of Miller's Cleveland data and Miller's own analysis of his Mt. Wilson data
Both Miller and I use a bandpass filter on each set of turns: that is, we reduce each set to second harmonic coefficients, i. e. an azimuth and amplitude for the second harmonic approximant. But Miller uses another filter which I haven't: for the unevenly distributed data over one turn of the Earth (at a given epoch spanning a few days) Miller not only replaces the azimuth & amplitude data with first harmonic approximants (i.e. period one day); he also subtracts the mean from the azimuth. However, I tried this by subtracting the unweighted mean coefficient of sin(2*theta), and found about the same best-fitting drift direction, RA,Decl = 194,84, hardly any different from what I found without this reduction to mean zero.
Both Miller and I use a bandpass filter on each set of turns: that is, we reduce each set to second harmonic coefficients, i. e. an azimuth and amplitude for the second harmonic approximant. But Miller uses another filter which I haven't: for the unevenly distributed data over one turn of the Earth (at a given epoch spanning a few days) Miller not only replaces the azimuth & amplitude data with first harmonic approximants (i.e. period one day); he also subtracts the mean from the azimuth. However, I tried this by subtracting the unweighted mean coefficient of sin(2*theta), and found about the same best-fitting drift direction, RA,Decl = 194,84, hardly any different from what I found without this reduction to mean zero.
Please Log in or Create an account to join the conversation.
- Joe Keller
- Offline
- Platinum Member
Less
More
- Thank you received: 0
11 years 3 months ago #13980
by Joe Keller
Replied by Joe Keller on topic Reply from
Dear Pierre Fuerxer, Thomas Goodey, James DeMeo,
Yes, it is easier for me to mail you the 3.5" standard floppy disk, than it is for me to "cut and paste" the program into an email. The reason is, that I cannot find any way to "cut and paste" more than a few lines each time, so I must do more than a thousand "mouse clicks". The data are column sums, because this results in a twenty-fold saving of transcription labor for me, as opposed to entering the row of data for every turn of the interferometer.
The program has my own analysis (with corrections of my previous errors) which some people might find useful. But also, the program contains the data for all of Miller's steel interferometer experiments in Cleveland. (Before, I only had the data from 1922-1924, but now it is all of the Cleveland steel interferometer data, 1905-1929.) There were 140 data sheets also known as "sets" or "experiments"; each data sheet resulted in a DATA line with 16 data for the interferometer readings at the 16 azimuths (column sums of the 20, more or less, turns per set) and on a DATA line above that, another 8 data giving the midpoint time of the set (5 data for date and time), number of turns in the set, Goodey-Keller page reference (the standard pagination of the definitive, i.e. the Case Archives / DeMeo, version of Miller's notebooks, which some Iowa State college students made under Goodey's supervision when Thomas visited me in Ames in Feb. 2004) and number of fringes in view (important for evaluating the Hicks effect). Also, the program contains "REM" (i.e. reminder) statements explaining the many interpretations and emendations I had to make: correction of addition errors in the columns, adjustment for varying numbers of mirrors or varying labeling of North, omission of control experiments using heaters, determination of the most correct time, etc.
In sum, the program is not only the data; it is annotated with the important emendations and explanations I gleaned from what Miller wrote on these pages, and also there is my own analysis program.
As the program exists now, my detrending procedure still is erroneous. I believe I now understand how to detrend correctly, but my preliminary estimate of the effect of detrending, is that it will affect the result by less than one degree. I ran the program without detrending (coarse search grid only, on my slow "Intel 486" home computer) and found that on a 10x10 degree grid, my erroneous detrending made no difference in the grand result, vs. no detrending. So, my latest grand result, below, for the Cleveland drift vector, which I also have posted on the messageboard of www.metaresearch.org , should not differ significantly, that is to say should not differ more than 10 degrees, from the properly detrended result. It is possible that some of the subgroups, say "1922" for example, might be significantly affected by correct detrending.
Latest version, posted yesterday to metaresearch.org:
The best fitting (least sum of squared differences of observed and predicted second harmonic coefficients for each set) drift directions (RA, Decl) with correlation coefficient of observed & expected, sigma, apparent drift speed, no. of sets, no. of turns, and mean Julian Date weighted by no. of turns:
1905* (one July set, four Oct-Nov sets)
RA,Decl = 210, 25
+0.74,+2.52,11.6km/s
5,230,2417119.880
Apr 1922
0,85
+0.68,+2.19,34.8km/s
5,81,2423162.415
Aug-Sep 1923
15,30
+0.44,+2.73,12.6km/s
18,263,2423660.527
Jun-Jul 1924
90,80
+0.50,+4.93,12.3km/s
41,608,2423979.216
Apr 1927
(There was a (-) correlation that was larger, -0.34.)
265,45
+0.26,+1.62,7.2km/s
20,400,2424984.541
Aug 1927
90,80
+0.45,+4.22,10.3km/s
40,800,2425110.636
Sep-Oct 1929
220,65
+0.67,+3.51,13.0km/s
11,220,2425887.935
Grand result for all data, 1905-1929, for the steel interferometer at Cleveland:
RA, Decl = 196,82
corr coeff = +0.2392, sigma = +4.060, apparent drift speed = 9.945km/s
140 sets, 2602 turns, mean Julian Date 2423979.047 weighted by turns
* For the data subsets, I made only a coarse search on a 5x5deg northern hemisphere (including equator) grid with rough sec(Decl) correction in RA steps. For the grand data set, this was followed by strict 1x1deg search on the largest grid square centered on the best coarse point and not overlapping other coarse points.
I had reason to doubt the sign of the 1905 data (ambiguous notation in Miller's notes, giving me two possible interpretations) so I chose the interpretation for 1905 giving the largest positive correlation. As noted above, I could have obtained a larger positive correlation by reversing the sign of the Apr 1927 data but lacking any other reason, did not do so. Likewise although a convenient way to explain the three times greater apparent drift speed in 1922 would be to assume that Miller was recording hundredths, not tenths, of a fringe that year, I did not make this emendation either, because I find no evidence for it on that year's data sheets.
The coarse search on the grand data set found sigma = +4.032 for RA,Decl = 210,80. So, the significance, p, was almost as great, for a direction three great-circle degrees away from the best. This suggests, by extrapolation of sigma and integration weighted by 1/p, an error bar of at least 10 degrees, even for the grand result.
The Hicks first harmonic correlation was much larger, when I assumed that the Hicks effect should be 90deg out of phase with the drift azimuth, as I believe Hicks did also. If the Hicks effect leads by 90deg, then the largest correlation coefficient is +0.15, sigma = +2.55, at RA,Decl = 339,71; this is 26 deg from the best second harmonic indicated direction given above.
- Joe Keller
Yes, it is easier for me to mail you the 3.5" standard floppy disk, than it is for me to "cut and paste" the program into an email. The reason is, that I cannot find any way to "cut and paste" more than a few lines each time, so I must do more than a thousand "mouse clicks". The data are column sums, because this results in a twenty-fold saving of transcription labor for me, as opposed to entering the row of data for every turn of the interferometer.
The program has my own analysis (with corrections of my previous errors) which some people might find useful. But also, the program contains the data for all of Miller's steel interferometer experiments in Cleveland. (Before, I only had the data from 1922-1924, but now it is all of the Cleveland steel interferometer data, 1905-1929.) There were 140 data sheets also known as "sets" or "experiments"; each data sheet resulted in a DATA line with 16 data for the interferometer readings at the 16 azimuths (column sums of the 20, more or less, turns per set) and on a DATA line above that, another 8 data giving the midpoint time of the set (5 data for date and time), number of turns in the set, Goodey-Keller page reference (the standard pagination of the definitive, i.e. the Case Archives / DeMeo, version of Miller's notebooks, which some Iowa State college students made under Goodey's supervision when Thomas visited me in Ames in Feb. 2004) and number of fringes in view (important for evaluating the Hicks effect). Also, the program contains "REM" (i.e. reminder) statements explaining the many interpretations and emendations I had to make: correction of addition errors in the columns, adjustment for varying numbers of mirrors or varying labeling of North, omission of control experiments using heaters, determination of the most correct time, etc.
In sum, the program is not only the data; it is annotated with the important emendations and explanations I gleaned from what Miller wrote on these pages, and also there is my own analysis program.
As the program exists now, my detrending procedure still is erroneous. I believe I now understand how to detrend correctly, but my preliminary estimate of the effect of detrending, is that it will affect the result by less than one degree. I ran the program without detrending (coarse search grid only, on my slow "Intel 486" home computer) and found that on a 10x10 degree grid, my erroneous detrending made no difference in the grand result, vs. no detrending. So, my latest grand result, below, for the Cleveland drift vector, which I also have posted on the messageboard of www.metaresearch.org , should not differ significantly, that is to say should not differ more than 10 degrees, from the properly detrended result. It is possible that some of the subgroups, say "1922" for example, might be significantly affected by correct detrending.
Latest version, posted yesterday to metaresearch.org:
The best fitting (least sum of squared differences of observed and predicted second harmonic coefficients for each set) drift directions (RA, Decl) with correlation coefficient of observed & expected, sigma, apparent drift speed, no. of sets, no. of turns, and mean Julian Date weighted by no. of turns:
1905* (one July set, four Oct-Nov sets)
RA,Decl = 210, 25
+0.74,+2.52,11.6km/s
5,230,2417119.880
Apr 1922
0,85
+0.68,+2.19,34.8km/s
5,81,2423162.415
Aug-Sep 1923
15,30
+0.44,+2.73,12.6km/s
18,263,2423660.527
Jun-Jul 1924
90,80
+0.50,+4.93,12.3km/s
41,608,2423979.216
Apr 1927
(There was a (-) correlation that was larger, -0.34.)
265,45
+0.26,+1.62,7.2km/s
20,400,2424984.541
Aug 1927
90,80
+0.45,+4.22,10.3km/s
40,800,2425110.636
Sep-Oct 1929
220,65
+0.67,+3.51,13.0km/s
11,220,2425887.935
Grand result for all data, 1905-1929, for the steel interferometer at Cleveland:
RA, Decl = 196,82
corr coeff = +0.2392, sigma = +4.060, apparent drift speed = 9.945km/s
140 sets, 2602 turns, mean Julian Date 2423979.047 weighted by turns
* For the data subsets, I made only a coarse search on a 5x5deg northern hemisphere (including equator) grid with rough sec(Decl) correction in RA steps. For the grand data set, this was followed by strict 1x1deg search on the largest grid square centered on the best coarse point and not overlapping other coarse points.
I had reason to doubt the sign of the 1905 data (ambiguous notation in Miller's notes, giving me two possible interpretations) so I chose the interpretation for 1905 giving the largest positive correlation. As noted above, I could have obtained a larger positive correlation by reversing the sign of the Apr 1927 data but lacking any other reason, did not do so. Likewise although a convenient way to explain the three times greater apparent drift speed in 1922 would be to assume that Miller was recording hundredths, not tenths, of a fringe that year, I did not make this emendation either, because I find no evidence for it on that year's data sheets.
The coarse search on the grand data set found sigma = +4.032 for RA,Decl = 210,80. So, the significance, p, was almost as great, for a direction three great-circle degrees away from the best. This suggests, by extrapolation of sigma and integration weighted by 1/p, an error bar of at least 10 degrees, even for the grand result.
The Hicks first harmonic correlation was much larger, when I assumed that the Hicks effect should be 90deg out of phase with the drift azimuth, as I believe Hicks did also. If the Hicks effect leads by 90deg, then the largest correlation coefficient is +0.15, sigma = +2.55, at RA,Decl = 339,71; this is 26 deg from the best second harmonic indicated direction given above.
- Joe Keller
Please Log in or Create an account to join the conversation.
- Joe Keller
- Offline
- Platinum Member
Less
More
- Thank you received: 0
10 years 9 months ago #21986
by Joe Keller
Replied by Joe Keller on topic Reply from
Hi all!
I'm still here and I'm OK, despite the six month silence. I've been busy planning, buying (in Oct. 2013) and caring for two Gotland ponies. Also my Samoyed dam had seven live puppies in Dec. 2013. This and other agricultural enterprises have been keeping me too busy to think about astronomy much.
I recommend the recent three-part series in Sky & Telescope, about the structure of our Milky Way. It's excellent!
A few days ago I heard the former Governor of Iowa, Governor Vilsack, now U. S. Secretary of Agriculture, being interviewed on National Public Radio. He said that while he didn't like the term "global warming" (a way of saying that he didn't believe in global warming) he was sure that we were in an epoch of climate extremes, with unusually many extremes of high and low temp, wet and dry.
I'm still here and I'm OK, despite the six month silence. I've been busy planning, buying (in Oct. 2013) and caring for two Gotland ponies. Also my Samoyed dam had seven live puppies in Dec. 2013. This and other agricultural enterprises have been keeping me too busy to think about astronomy much.
I recommend the recent three-part series in Sky & Telescope, about the structure of our Milky Way. It's excellent!
A few days ago I heard the former Governor of Iowa, Governor Vilsack, now U. S. Secretary of Agriculture, being interviewed on National Public Radio. He said that while he didn't like the term "global warming" (a way of saying that he didn't believe in global warming) he was sure that we were in an epoch of climate extremes, with unusually many extremes of high and low temp, wet and dry.
Please Log in or Create an account to join the conversation.
- Larry Burford
- Offline
- Platinum Member
Less
More
- Thank you received: 0
10 years 9 months ago #21987
by Larry Burford
Replied by Larry Burford on topic Reply from Larry Burford
<b>[Joe Keller] "... an epoch of climate extremes, with unusually many extremes of high and low temp, wet and dry."</b>
Climate Change.
The (New/Next) Ice Age.
The (New/Next) Global Warming.
***
Democratic/Liberal pundits say "it" is a done deal, already too late to fix (but even if it isn't (a done deal), we ought to fix it anyway).
Republican/Conservative pundits say it isn't happening (and even if it is, it isn't being caused by human actions.)
<ul>Have you ever noticed that when political factions argue with each other, they never use the same "buzz words" (never talk about the same things)? How else can they keep their idiotic followers hooting?</ul>
Roughly a quarter of the population sends regular checks to the politicians on the conservative side of this (or any) """""debate""""".
At the same time, roughly a quarter of the population (NOT the same quarter, I hope you see that) sends regular checks to the politicians on the liberal side of this (or any) """""debate""""".
The rest of us pretty much ignore the *ssh*l*s in the aforementioned groups. The problem with this tactic is that while you can ignore them, they (the libs and the cons) are NOT ignoring you. Your paycheck (if you are an employee) or your profits (if you are an employer) or your benefits (if you are a recipient) are clearly visible on their radar.
And since they are the government they have access to your money.
***
So, IMO climate change exists whether or not it is caused by/exacerbated by human action.
Climate change has always existed. We can clearly see it in the geological record. Why should now be any different?
Oh yeah - the contribution of human activity. (Only *NOW* has this become important.) IMO, it is not credible to claim that human activity has had NO IMPACT on climate change. But just as obviously it is not credible to claim that it is the only reason we have climate change right now.
We have been using energy at an exponentially increasing rate for thousands of years. Sooner or later this is going to tip a balance somewhere.
***
Is this a Bad Thing?
Or is it a Good Thing?
Or is it just a thing?
Your opinion is solicited.
***
Man-made or natural or some combination of the two, if our climate deviates too far off the norm we are SERIOUSLY screwed. As in the proverbial 'extinction level event'.
***
Should we start a new forum to talk about this? It seems a bit off topic here.
Climate Change.
The (New/Next) Ice Age.
The (New/Next) Global Warming.
***
Democratic/Liberal pundits say "it" is a done deal, already too late to fix (but even if it isn't (a done deal), we ought to fix it anyway).
Republican/Conservative pundits say it isn't happening (and even if it is, it isn't being caused by human actions.)
<ul>Have you ever noticed that when political factions argue with each other, they never use the same "buzz words" (never talk about the same things)? How else can they keep their idiotic followers hooting?</ul>
Roughly a quarter of the population sends regular checks to the politicians on the conservative side of this (or any) """""debate""""".
At the same time, roughly a quarter of the population (NOT the same quarter, I hope you see that) sends regular checks to the politicians on the liberal side of this (or any) """""debate""""".
The rest of us pretty much ignore the *ssh*l*s in the aforementioned groups. The problem with this tactic is that while you can ignore them, they (the libs and the cons) are NOT ignoring you. Your paycheck (if you are an employee) or your profits (if you are an employer) or your benefits (if you are a recipient) are clearly visible on their radar.
And since they are the government they have access to your money.
***
So, IMO climate change exists whether or not it is caused by/exacerbated by human action.
Climate change has always existed. We can clearly see it in the geological record. Why should now be any different?
Oh yeah - the contribution of human activity. (Only *NOW* has this become important.) IMO, it is not credible to claim that human activity has had NO IMPACT on climate change. But just as obviously it is not credible to claim that it is the only reason we have climate change right now.
We have been using energy at an exponentially increasing rate for thousands of years. Sooner or later this is going to tip a balance somewhere.
***
Is this a Bad Thing?
Or is it a Good Thing?
Or is it just a thing?
Your opinion is solicited.
***
Man-made or natural or some combination of the two, if our climate deviates too far off the norm we are SERIOUSLY screwed. As in the proverbial 'extinction level event'.
***
Should we start a new forum to talk about this? It seems a bit off topic here.
Please Log in or Create an account to join the conversation.
10 years 9 months ago #22382
by Jim
Replied by Jim on topic Reply from
Climate change is an important topic not relative to anything at this site. It might be better to do what you did with political matters and get it out of this forum that DR Joe has owned for years.
Please Log in or Create an account to join the conversation.
Time to create page: 0.422 seconds