*

Offline Pete Svarrior

  • e
  • Planar Moderator
  • *****
  • Posts: 16073
  • (◕˽ ◕ ✿)
    • View Profile
Re: Rowbotham experiment #9
« Reply #40 on: January 04, 2019, 06:30:22 PM »
You are inserting irrelevancy by asking him to prove why 0.05% of the earth is statistically significant for the whole, when that is not what he was setting out to do.
He is demanding that we explain a tiny subset of the data. I ask why he's focusing on that tiny subset. I do not want to assume his reasoning, hence the question.

If you really wanted me to guess a reason, I would speculate confirmation bias - the tiny speck on the map confirms his hypothesis, and nothing else does, so let's focus on favourable data. I'd rather have him explain himself than assume that - I think that's only courteous.
Read the FAQ before asking your question - chances are we already addressed it.
Follow the Flat Earth Society on Twitter and Facebook!

If we are not speculating then we must assume

Offline edby

  • *
  • Posts: 1214
    • View Profile
Re: Rowbotham experiment #9
« Reply #41 on: January 04, 2019, 07:36:07 PM »
You are inserting irrelevancy by asking him to prove why 0.05% of the earth is statistically significant for the whole, when that is not what he was setting out to do.
He is demanding that we explain a tiny subset of the data. I ask why he's focusing on that tiny subset. I do not want to assume his reasoning, hence the question.

If you really wanted me to guess a reason, I would speculate confirmation bias - the tiny speck on the map confirms his hypothesis, and nothing else does, so let's focus on favourable data. I'd rather have him explain himself than assume that - I think that's only courteous.
I already explained this in my earlier post. As I live in England, I am interested in whether England is flat or not. Evidence from Findlay's data strongly suggests not flat. Rest of world, not too bothered.

*

Offline Pete Svarrior

  • e
  • Planar Moderator
  • *****
  • Posts: 16073
  • (◕˽ ◕ ✿)
    • View Profile
Re: Rowbotham experiment #9
« Reply #42 on: January 04, 2019, 07:49:53 PM »
I already explained this in my earlier post. As I live in England, I am interested in whether England is flat or not. Evidence from Findlay's data strongly suggests not flat. Rest of world, not too bothered.
That's extremely peculiar. At such small scale, you can only really hope to infer how hilly the area is. Let's say it turns out that England is slightly convex - how does that help this particular discussion?

And, if it doesn't, I hope you can guess what I have to say next.
Read the FAQ before asking your question - chances are we already addressed it.
Follow the Flat Earth Society on Twitter and Facebook!

If we are not speculating then we must assume

Offline edby

  • *
  • Posts: 1214
    • View Profile
Re: Rowbotham experiment #9
« Reply #43 on: January 04, 2019, 10:14:44 PM »
I already explained this in my earlier post. As I live in England, I am interested in whether England is flat or not. Evidence from Findlay's data strongly suggests not flat. Rest of world, not too bothered.
That's extremely peculiar. At such small scale, you can only really hope to infer how hilly the area is. Let's say it turns out that England is slightly convex - how does that help this particular discussion?
Remember we are talking about a wide coastal area around England, i.e. about the sea being 'hilly', or rather having a consistent positive curvature. That would be significant, no? Of course this could be a peculiarity of the English coastline, but would be something FE needed to explain.

Offline edby

  • *
  • Posts: 1214
    • View Profile
Re: Rowbotham experiment #9
« Reply #44 on: January 05, 2019, 03:32:58 PM »
I have just checked out the 1879 edition of A Description and List of the Lighthouses of the World where some of the numbers have changed. Mostly they change the underestimates of visible distance to something closer to the expected value. Sometimes they change overestimates to something close. Very occasionally they make the discrepancies worse.

My point therefore remains that this work (on which Rowbotham relied) is not entirely reliable.

There is also an 1887 edition, which is not available online, but is in the library and I shall check out when next there.

Offline edby

  • *
  • Posts: 1214
    • View Profile
Re: Rowbotham experiment #9
« Reply #45 on: January 05, 2019, 04:43:20 PM »
Further, the 1879 edition states

Quote
The distance to which the principal lights are visible is generally limited by the horizon. There is no doubt but that they might be seen to very great distances, even 60, 80, or even 100 miles, if sufficient elevation could be gained to view them from. It is considered by many that 250 feet is the maximum height necessary or advisable, which will give an horizon 18 miles distant and, by ascending the rigging, 20 miles off.


adding that the table gives miles visible to an observer 14 feet above the sea. It then gives a table 'given in the grand work on the Skerryvore Lighthouse by Alan Stevenson esq, which lists the distance to the horizon in both statute and nautical miles'.

I have calibrated the table and find it gives exact results using the formula d = 1.3228 x sqrt(h), where d is distance to horizon, h is height of observer.

What do we conclude? Well first that Findlay and co were assuming a spherical earth in their calculations. Second, that it is odd that Rowbotham was using their book as proof of a flat earth. The logical conclusion is that the exceptions that Rowbotham he found were simply typographical or reporting or calculation errors, some of which were corrected in the 1879 edition, some were not.

[EDIT]For those who are interested, the book by Stevenson to which he refers is here https://books.google.co.uk/books?id=u-hhAAAAcAAJ, and the table is on p.329. Stevenson mentions the formula d = sqrt(7h)/2, and you will notice that sqrt(7)/2 = 1.3228, the number I calibrated above.

You may also remember that Robert Louis Stevenson, the Treasure Island guy, Captain Hook and all that, came from a family of lighthouse engineers. He was the nephew of Alan. He wasn't so good at lighthouses, so he wrote stories instead.
« Last Edit: January 05, 2019, 04:55:54 PM by edby »

*

Offline Pete Svarrior

  • e
  • Planar Moderator
  • *****
  • Posts: 16073
  • (◕˽ ◕ ✿)
    • View Profile
Re: Rowbotham experiment #9
« Reply #46 on: January 05, 2019, 04:58:31 PM »
Of course this could be a peculiarity of the English coastline, but would be something FE needed to explain.
Only if we assume the data is correct. Given that a vast majority of it is not, I see no reason to make that assumption. You're conflating "correlation with RET" with "correctness".
Read the FAQ before asking your question - chances are we already addressed it.
Follow the Flat Earth Society on Twitter and Facebook!

If we are not speculating then we must assume

Offline edby

  • *
  • Posts: 1214
    • View Profile
Re: Rowbotham experiment #9
« Reply #47 on: January 05, 2019, 05:12:54 PM »
Of course this could be a peculiarity of the English coastline, but would be something FE needed to explain.
Only if we assume the data is correct. Given that a vast majority of it is not, I see no reason to make that assumption. You're conflating "correlation with RET" with "correctness".
I agree there are problems with the data. See my later posts immediately above.

However correlation still needs explanation, and there I disagree with you. Correlation implies lack of randomness, and lack of randomness always requires explanation.

[EDIT] You also mentioned a 0.05% figure somewhere above. Where did this come from?

[EDIT] Oh I see. You divided area of UK by area of world (assuming latter is correct) giving 0.05%. But immediately above you say "Given that a vast majority of it [i.e. the data] is not". But the percentage of Findlay's book devoted to UK data is not 0.05%. Oh dear.
« Last Edit: January 05, 2019, 05:18:04 PM by edby »

Offline edby

  • *
  • Posts: 1214
    • View Profile
Re: Rowbotham experiment #9
« Reply #48 on: January 05, 2019, 05:44:57 PM »
In fact 11% of Findlay's data is devoted to UK (i.e. England and Scotland) lighthouses.

*

Offline Pete Svarrior

  • e
  • Planar Moderator
  • *****
  • Posts: 16073
  • (◕˽ ◕ ✿)
    • View Profile
Re: Rowbotham experiment #9
« Reply #49 on: January 05, 2019, 05:48:30 PM »
Correlation implies lack of randomness, and lack of randomness always requires explanation.
2*2=2+2, therefore for a certain subset of numbers, addition is the same as multiplication. Those who propose that these operations are different must be able to explain this correlation, even though it's not in any way significant or remarkable.

Correlation is a perfectly valid outcome of randomness or coincidence. Indeed, it's exactly as likely as any other outcome. You're attempting to force a logical fallacy through.
« Last Edit: January 05, 2019, 05:51:00 PM by Pete Svarrior »
Read the FAQ before asking your question - chances are we already addressed it.
Follow the Flat Earth Society on Twitter and Facebook!

If we are not speculating then we must assume

*

Offline Pete Svarrior

  • e
  • Planar Moderator
  • *****
  • Posts: 16073
  • (◕˽ ◕ ✿)
    • View Profile
Re: Rowbotham experiment #9
« Reply #50 on: January 05, 2019, 05:53:14 PM »
As for 0.05%, I said this:

You've found a correlation that allegedly applies to >0.05% of the Earth.
Of course, I meant <0.05%, but hey-ho, this isn't the first time I've failed at basic typing ;)

Nonetheless, it was you who claimed that your data shows the UK is convex while the rest of the world is not. I can't be held responsible for your own claims.
« Last Edit: January 05, 2019, 05:58:54 PM by Pete Svarrior »
Read the FAQ before asking your question - chances are we already addressed it.
Follow the Flat Earth Society on Twitter and Facebook!

If we are not speculating then we must assume

Offline edby

  • *
  • Posts: 1214
    • View Profile
Re: Rowbotham experiment #9
« Reply #51 on: January 05, 2019, 06:07:28 PM »
My claims were about data, not area. See this post https://forum.tfes.org/index.php?topic=11728.msg178123#msg178123.
Also this post https://forum.tfes.org/index.php?topic=11728.msg178183#msg178183.

You say
He is demanding that we explain a tiny subset of the data.
But it is not a ‘tiny subset of the data’, as I have explained. You are confusing percentage area, by percentage of data. Keep on digging that hole.

2*2=2+2, therefore for a certain subset of numbers, addition is the same as multiplication.
Yes for a very small data set, in this case one, correlation is of no statistical significance whatsoever. For a large set, in my case I used 54 items, it is highly significant. Another way to approach it is via the concept of ‘standard deviation’. How far does the expected amount (RE calculation) differ from the observation? In the case of Findlay’s UK data, 1.67 miles.

*

Offline Pete Svarrior

  • e
  • Planar Moderator
  • *****
  • Posts: 16073
  • (◕˽ ◕ ✿)
    • View Profile
Re: Rowbotham experiment #9
« Reply #52 on: January 05, 2019, 11:41:43 PM »
You keep asserting that it's significant with no qualification. I have already asked you to substantiate that. Further unqualified assertions are unlikely to advance your cause.

Also, if you want to shift the goalposts to percentage of data points (despite the obvious flaws of that approach), even after you've successfully made your case you'll still need to justify your reasoning for ignoring the vast majority of data points you personally find inconvenient. Whether you fight against 99.5% of the Earth's area or 90% of the data points collected is largely irrelevant - it's nothing but meaningless semantics that ignores the actual issue.

Start defending your position.
« Last Edit: January 05, 2019, 11:52:33 PM by Pete Svarrior »
Read the FAQ before asking your question - chances are we already addressed it.
Follow the Flat Earth Society on Twitter and Facebook!

If we are not speculating then we must assume

*

Offline stack

  • *
  • Posts: 3583
    • View Profile
Re: Rowbotham experiment #9
« Reply #53 on: January 06, 2019, 02:04:53 AM »
You keep asserting that it's significant with no qualification. I have already asked you to substantiate that. Further unqualified assertions are unlikely to advance your cause.

Also, if you want to shift the goalposts to percentage of data points (despite the obvious flaws of that approach), even after you've successfully made your case you'll still need to justify your reasoning for ignoring the vast majority of data points you personally find inconvenient. Whether you fight against 99.5% of the Earth's area or 90% of the data points collected is largely irrelevant - it's nothing but meaningless semantics that ignores the actual issue.

This is essentially what Bresher's point was in describing what Rowbotham did. Bresher claimed that, after reviewing the reference source, Rowbotham, was, in your words, 'ignoring the vast majority of data points he personally found inconvenient.'

Offline edby

  • *
  • Posts: 1214
    • View Profile
Re: Rowbotham experiment #9
« Reply #54 on: January 06, 2019, 10:17:25 AM »
[…] if you want to shift the goalposts to percentage of data points
I already stated above, citing two previous posts, that ‘my claims were about data, not area’. So I am not ‘shifting goalposts’.  The whole of this thread is about the data that Rowbotham and others have used to support their claim of a flat earth, all of which (as far as I can tell) is sourced from Findlay 1862.

Quote
You keep asserting that it's significant with no qualification. I have already asked you to substantiate that.
You did, and I replied citing this as evidence. Or look in almost any standard statistics textbook that deals with correlation.  [EDIT] I used 54 data points for the UK. Most textbooks state that 20 will suffice.

I will deal with the allegations of cherry-picking in the next post.
« Last Edit: January 06, 2019, 10:23:14 AM by edby »

Offline edby

  • *
  • Posts: 1214
    • View Profile
Re: Rowbotham experiment #9
« Reply #55 on: January 06, 2019, 10:19:19 AM »
… even after you've successfully made your case you'll still need to justify your reasoning for ignoring the vast majority of data points you personally find inconvenient. Whether you fight against 99.5% of the Earth's area or 90% of the data points collected is largely irrelevant - it's nothing but meaningless semantics that ignores the actual issue.
Start defending your position.

My position is that there are many clear problems with the data in the 1862 edition of the book ‘Lighthouses of the World’, which Samuel Rowbotham cites in Earth not a Globe, experiment number 9.

First problem. Rowbotham cites Findlay as defining the visible range as the minimum distance to which the light can be seen in clear weather from a height of 10 feet above the sea level’. This is what the 1862 ed. states, p. 32. However the 1879 edition p.31 states ‘The distance of the sea-horizon due to that elevation [of the lighthouse], is stated to be the distance it may be seen from the deck of an ordinary vessel, 14 feet above the sea’

This discrepancy affects every calculation in Findlay’s book.

Second problem A number of figures given in the 1862 edition were revised in the 1879 edition, some of them cited by Rowbotham. For example, Rowbothamsays ‘The Light on Cape Bonavista, Newfoundland, is 150 feet above high water, and is visible 35 statute miles [=30 NM]. These figures will give, on calculating for the earth's rotundity, 491 feet as the distance it should be sunk below the sea horizon’. This calculation is correct assuming the figures given in the 1862 ed. p.111. Howeverthe 1879 ed., p. 157, gives 16 miles, not 30. If we assume the observer is 16 feet above water, not 10, and assume nautical miles are meant (Findlay never says), this is less (not more) than the RE calculation.

Rowbotham writes ‘The Port Nicholson Light, in New Zealand (erected in 1859), is visible 35 statute miles, the altitude being 420 feet above high water. If the water is convex it ought to be 220 feet below the horizon’. But the 1879 edition (p.155) gives a height of 450.

Of course the 1879 edition could be wrong and the 1862 correct, but this contradicts the usual state of things where later editions correct mistakes in earlier ones, and in any case it casts doubt on the reliability of Findlay’s data, on which Rowbotham’s claims depend.

Third problemWhen we split Findlay’s data into UK (11%) and non-UK (89%) data, we find a significant correlation (97.5%) between RE and Findlay UK estimate of visible range, and a very low (24%) correlation for non-UK.

Pete has challenged my claim of ‘significant’ for the 97.5% correlation. This is for over 50 data points, I refer him to any standard textbook on statistics on this point.

He has also claimed I am cherry picking the data. But my position, as stated above, is that there are clear problems with the data that Rowbotham was using (i.e. Findlay’s 1862 data). Either

(1)  All the data, UK and non-UK, is weak. Then I have my case. Rowbotham was basing his claims on weak data.

(2) Some of the data is weak (the non-UK data), other data is strong (the UK data). This is my hypothesis. Findlay was almost certainly relying on agencies for the non-UK data, and weak data is what you often get with agencies. But if the UK data is strong, that supports the RE position, moreover Rowbotham’s case collapses.

(3) The non-UK data is strong, the UK data is weak. This contradicts the strong correlation found in the UK case. For the non-UK case we have merely weak correlation, which proves nothing.

(4) Both data sets are strong. But in that case we have the sea around the UK being convex, the sea around all other countries flat. Pete has talked about the UK being ‘hilly’, but the last time I looked, hills only exist on solid ground. We are talking here about the surface of the sea here, not the land, and the only ‘hills’ on the sea are temporary ones. I suppose it could be argued that there is a sort of permanent swell surrounding the British coastline, but that is implausible even in an FE state of mind.

I rest my case.

*

Offline Pete Svarrior

  • e
  • Planar Moderator
  • *****
  • Posts: 16073
  • (◕˽ ◕ ✿)
    • View Profile
Re: Rowbotham experiment #9
« Reply #56 on: January 06, 2019, 12:39:55 PM »
(3) The non-UK data is strong, the UK data is weak. This contradicts the strong correlation found in the UK case. For the non-UK case we have merely weak correlation, which proves nothing.
Your personal credulity is not applicable here. Correlations prove nothing, and your attempt to categorise them into correlations you like and correlations you dislike is laughable.

I rest my case.
I take it we're ready for the AR move, then? Your entire argument relies on correlations implying causation, and on the fact that you personally think some things are likely. A perfect exhibit for the Logic Museum, indeed.
Read the FAQ before asking your question - chances are we already addressed it.
Follow the Flat Earth Society on Twitter and Facebook!

If we are not speculating then we must assume

Offline edby

  • *
  • Posts: 1214
    • View Profile
Re: Rowbotham experiment #9
« Reply #57 on: January 06, 2019, 01:20:59 PM »
(3) The non-UK data is strong, the UK data is weak. This contradicts the strong correlation found in the UK case. For the non-UK case we have merely weak correlation, which proves nothing.
Your personal credulity is not applicable here. Correlations prove nothing, and your attempt to categorise them into correlations you like and correlations you dislike is laughable.
As I said, look at any textbook entry that deals with data size and correlation. Not my 'personal credulity'.

*

Offline Pete Svarrior

  • e
  • Planar Moderator
  • *****
  • Posts: 16073
  • (◕˽ ◕ ✿)
    • View Profile
Re: Rowbotham experiment #9
« Reply #58 on: January 06, 2019, 01:27:22 PM »
As I said - trying to leap from correlation to a causal relation is laughable. Revise your argument or justify it.
Read the FAQ before asking your question - chances are we already addressed it.
Follow the Flat Earth Society on Twitter and Facebook!

If we are not speculating then we must assume

Offline edby

  • *
  • Posts: 1214
    • View Profile
Re: Rowbotham experiment #9
« Reply #59 on: January 06, 2019, 01:34:16 PM »
As I said - trying to leap from correlation to a causal relation is laughable. Revise your argument or justify it.
Nowhere in any of this thread did I use the word 'causation' or its cognates. I said that any high correlation requires an explanation, for correlation implies a degree of non-randomness.

You are the only person talking about causal relations here.