welcome to the fest
 
 FAQFAQ   SearchSearch   MemberlistMemberlist   UsergroupsUsergroups   RegisterRegister 
 ProfileProfile   Log in to check your private messagesLog in to check your private messages   Log inLog in 

Sept. 6: Veggie detector!
Goto page Previous  1, 2, 3 ... , 23, 24, 25  Next
 
Post new topic   Reply to topic    Sinfest Forum Index -> Sinfest
View previous topic :: View next topic  
Author Message
Dogen



Joined: 10 Jul 2006
Posts: 10798
Location: Bellingham, WA

PostPosted: Sat Sep 15, 2012 6:12 pm    Post subject: Reply with quote

I think the idea that no one is going to change their mind is weird, especially for someone who claims to work in science. I like being challenged. Sure, I argue, but my positions aren't set in stone, nor should they be. I follow the evidence (which is why I'm right in this), not my gut or tradition.

Edit: and when the evidence is contradictory I evaluate the strength of the evidence, and if it's equivalent or nearly equivalent I build that uncertainty into my opinion and wait for more. That just seems rational.
_________________
"Worse comes to worst, my people come first, but my tribe lives on every country on earth. Iíll do anything to protect them from hurt, the human race is what I serve." - Baba Brinkman


Last edited by Dogen on Sat Sep 15, 2012 6:20 pm; edited 1 time in total
Back to top
View user's profile Send private message
Samsally



Joined: 10 Jul 2006
Posts: 6508

PostPosted: Sat Sep 15, 2012 6:20 pm    Post subject: Reply with quote

Yeah, no lies, I lose all will to have anything to do with someone if they refuse to even entertain the notion that they can be wrong sometimes.

Fun fact: This is exactly why there is a person in this world who thinks I hate them because of ponies. I don't, but it isn't worth trying to clear up the misunderstanding.
_________________
Samsally the GrayAce
Back to top
View user's profile Send private message
ShadowCell



Joined: 03 Aug 2008
Posts: 6080
Location: California

PostPosted: Sat Sep 15, 2012 6:45 pm    Post subject: Reply with quote

my favorite part is when they pull that card only after six or seven pages of heated argument in defense of the opinions that they now really don't care about strangely enough
Back to top
View user's profile Send private message
ChastMastr



Joined: 15 Jul 2012
Posts: 473
Location: Tampa Bay, Florida, US

PostPosted: Sat Sep 15, 2012 7:25 pm    Post subject: Reply with quote

Heretical Rants wrote:
And so, our heroes reached Page Twenty-Three. They had faced many hardships, but they had finally decided to put their differences behind them and stop talking about things. Their situation might have been looking up, but little did they realize that the long, heated debate about internet debates was nigh upon them. HUAHAHAHAHAHAAAAA



AHHHH! RECURSION! MY WIDDLE BWAIN! AAAHHHH *BOOM*
Back to top
View user's profile Send private message Visit poster's website
Heretical Rants



Joined: 21 Jul 2009
Posts: 5344
Location: No.

PostPosted: Sat Sep 15, 2012 7:47 pm    Post subject: Reply with quote

I like recursion!

Code:
internetDebate :: (Argument -> Argument) -> [Argument] -> [Argument]
internetDebate argue (x:xs)
                    |if-tedious x = internetDebate argue xs
                    |if-we-already-argued-about x = internetDebate rehash [x] -- regardless
                    |otherwise = internetDebate pickalick mostOffensive (internetDebate argue [x]) (internetDebate argue xs)
internetDebate startArgument [x] = First Thing You Can Think of To Argue About x
internetDebate okayImDoneRightAfterThis [] = Argument About Arguments

_________________
butts
Back to top
View user's profile Send private message
Sojobo



Joined: 12 Jul 2006
Posts: 2443

PostPosted: Sun Sep 16, 2012 4:08 am    Post subject: Reply with quote

Lich Mong wrote:
It's not "my minimum standard" is the one used by most fields for hypothesis testing.

Your quotation marks are inappropriate. I was not mocking the standard. In fact, you didn't even call it a minimum standard, I did. You are reacting against exactly nothing.

Lich Mong wrote:
You're allowed to say whatever you want.

The question is if you want it to be meaningful or not.

It is meaningful. Are you going to tell me that data that contradicts a hypothesis at a 90% confidence interval, but not a 95% is meaningless? Seriously, every time I ask you a question, you dodge rather than answering straight out. Answer this one. The entirety of our conversation is contained within this question. The implication is there. The conclusions drawn are still meaningful.

Lich Mong wrote:
My numbers has a statistical power of .1, which is much much better than most political polls you see, for example.

I didn't ask you if your example was better than most political polls, I asked whether you thought the numbers from the Council were as flimsy as yours.

Lich Mong wrote:
I would suspect they used a smaller % of the student population.

However, if they had larger numbers then they might have a smaller errors, depending on overall size. But, IDK what they had. They did not print errors and it's clearly not mean to be a scientific poll.

I didn't ask whether your numbers used a higher percentage of the population than the Council's numbers. I asked whether you thought their numbers were as flimsy as yours. Also, I didn't ask if you knew what numbers they used, I asked whether you think their numbers are anywhere near as flimsy as in your example. Do you?

Also, do you think it is honest to reply to a question about statistical sensitivity by only mentioning percentage of population used? The raw size of the sample is almost certainly more important (law of large numbers and all that). And we are all in agreement that the raw size of their sample was much, much larger than in your example, right?

Also2, I actually bet the Council used a higher than 10% portion of the population. You've still given no reason we should be assuming the numbers are based on a poll rather than actual compiled data. Enrollment and graduation are both numbers that are compiled and tracked, not gathered by polling.

Lich Mong wrote:
The number of fields is immaterial since each one is treated differently. For 1000 people each one would have a different error for each field based on the overall size of the feild.

The number of fields is not immaterial. The hypothesis we want to test is whether women are less successful in Engineering than they are in other fields in general. Data that demonstrate that the female success rate in Engineering is actually the highest of 10 fields is much more meaningful for concluding our hypothesis is false than data which only shows the rate to be higher than one other field.

Lich Mong wrote:
For your 1000 people example, I did a quick back-of-the-envelope calculation(which i still have even though my browser ate my post, thank God) and came up with a 35 student error for US graduate engineers, so....

What numbers are you using in your back-of-the-envelope calculation?

If I use your dropout rates, but assume 1000 people, evenly split, as you stated, I get standard deviations of:
Field A, boys: sqrt(.14*500*(1-.14)) = 7.76 students
Field A, girls: sqrt(.06*500*(1-.06)) = 5.31 students
Field B, boys: sqrt(.90*500*(1-.90)) = 6.71 students
Field B, girls: sqrt(.80*500*(1-.80)) = 8.94 students

The maximum possible standard deviation for a 500 student sample is only 11.18 students (using p = .50). I cannot find your number no matter what I try. Have you abandoned the binomial assumption you made in the previous post?

Anyway, these numbers are all statistically significant, as I said they would be.

Lich Mong wrote:
I guess this is the part you throw other numbers it might have been at me until I give you an error you like? I mean, it might have been 2000 or 127 or 13050; those are all numbers. Why don't you pick one until I tell you the error is 1% and you can claim that was the number you suspected all along.

First, this is just more wankery, and I should probably just ignore it, but I am not going to, though, because it's again so acutely hypocritical. You are the one who has been making up wild numbers that might make your points stick. I have, at every opportunity, claimed that the numbers are likely to be steady, exactly as you have agreed is probably the case.

Lich Mong wrote:
It depends on what they were trying to show. The poll was not scientific or they would have given errors.
Also, they clearly were not trying to show what you were trying to show or they would have simply polled dropout numbers.

Their methodology could still be right(and likely is) because they were not trying to show what you're trying to show.

Of course they weren't trying to show what I am showing. They are just publishing the data they've collected. I am saying that they wouldn't have published that data if they had only polled a ridiculously small and unrepresentative sample. Therefore, I can assume their numbers are fairly robust.

Lich Mong wrote:
Your methodology is wrong because you are trying to compare things that aren't comparable. The 2010 admission %'s are not comparable to the 2010 graduation %'s. You are comparing apples and oranges. You can't infer what you trying to infer from that poll.

You've told me you agree that the numbers are probably pretty steady. That makes the numbers comparable.

Lich Mong wrote:
Which makes me ask again why you don't go find the numbers you DO need.

I feel like I'm the only one doing research for this.

...

Why don't you go find the ones that do?

More numbers would probably make it even more clear that Ember was wrong, but my argument is not with Ember, it is with you. And my contention the whole time is that the numbers given already imply what I say they imply. I have no need for more numbers.

Lich Mong wrote:
I. have. in. every. post.
You can't infer what you want to infer from those numbers. I don't know how many more times I have to say it, how many more examples I have to give, or how much more research and numbers I have to crunch to get through to you.
Those. numbers. do. not. imply. what. you. want. them. to. imply.

Go ahead and answer my first question in this post. Tell me that no confidence level implies anything until you reach that magic number of 95%. Perhaps this all comes down to a difference of opinion on what "imply" means. We can sort all this out if you just straightforwardly answer the question.
_________________
"To love deeply in one direction makes us more loving in all others."
- Anne-Sophie Swetchine
Back to top
View user's profile Send private message
Adyon



Joined: 27 May 2012
Posts: 1163
Location: Behind my Cintiq

PostPosted: Sun Sep 16, 2012 6:37 am    Post subject: Reply with quote

I wonder if this thread can go on as long as the one with Kylra did? I tapped out earlier than Dogen though. He was a trooper in that thread the way the conversation kept coming back around. Will we get a soda break at page 30 here too?
_________________
Back to top
View user's profile Send private message Visit poster's website
Lich Mong



Joined: 31 May 2012
Posts: 475

PostPosted: Sun Sep 16, 2012 7:30 am    Post subject: Reply with quote

Wow, you think I would have learned the first time the browser ate it...
Sojobo wrote:

Your quotation marks are inappropriate. I was not mocking the standard. In fact, you didn't even call it a minimum standard, I did. You are reacting against exactly nothing.
I guess I should have just put quotes around "my." I was trying to highlight that it was not 'mine,' but the scientific method's. It was not meant to be an aggressive statement, but just one saying I was not pulling that number out of my own ass.
Sojobo wrote:

It is meaningful. Are you going to tell me that data that contradicts a hypothesis at a 90% confidence interval, but not a 95% is meaningless? Seriously, every time I ask you a question, you dodge rather than answering straight out. Answer this one. The entirety of our conversation is contained within this question. The implication is there. The conclusions drawn are still meaningful.
If you want your statement to have scientific weight, yes. But--again--it's not ME that's setting that bar. Additionally, one sigma is normally 68%, not 90%. But we don't even know if the 1% mark is even better then a 50/50 for that poll.

Now, look, I should not have made such a big deal about this to start. At the time in RL I was... well, let's just say with more distance I realize I should not have jumping down your throat as I did. The argument started innocuous enough, but I became more obsessed with it then I should have. People make wild unfounded claims all the time on the web(I'm no less innocent of that than anyone else); I should not have singled you out and just let yours pass.

Just wanted to get that off my chest. Confession is good for the soul.
Sojobo wrote:

I didn't ask you if your example was better than most political polls, I asked whether you thought the numbers from the Council were as flimsy as yours.
My numbers were not 'flimsy.' But, I explicitly stated I thought they used a lower statistical power for their poll.

Sojobo wrote:

I didn't ask whether your numbers used a higher percentage of the population than the Council's numbers. I asked whether you thought their numbers were as flimsy as yours. Also, I didn't ask if you knew what numbers they used, I asked whether you think their numbers are anywhere near as flimsy as in your example. Do you?
We clearly disagree on what 'flimsy' means in this context. What are you using it to mean? That they cared more about Type I errors than Type II? That they cared more about Type II errors then Type I? What? There are lots of ways to 'rate' polls.
Sojobo wrote:

Also, do you think it is honest to reply to a question about statistical sensitivity by only mentioning percentage of population used? The raw size of the sample is almost certainly more important (law of large numbers and all that). And we are all in agreement that the raw size of their sample was much, much larger than in your example, right?
Its likely they polled more than 100 people, yes. But, what matters is they're dealing with a larger overall population. Lots of people go to gradschool each year.
Sojobo wrote:

Also2, I actually bet the Council used a higher than 10% portion of the population. You've still given no reason we should be assuming the numbers are based on a poll rather than actual compiled data. Enrollment and graduation are both numbers that are compiled and tracked, not gathered by polling.
I'll bet you my house against yours they did not. You don't normally use a high statistical power for those kind of polls. They're for news papers, not journals.

Sojobo wrote:

The number of fields is not immaterial. The hypothesis we want to test is whether women are less successful in Engineering than they are in other fields in general. Data that demonstrate that the female success rate in Engineering is actually the highest of 10 fields is much more meaningful for concluding our hypothesis is false than data which only shows the rate to be higher than one other field.
Let me restate. If you want to say a larger % of girls dropped out of Engineering gradschool than guys, then what happens in the other fields doesn't matter.
That was your original claim.

Sojobo wrote:

What numbers are you using in your back-of-the-envelope calculation?

If I use your dropout rates, but assume 1000 people, evenly split, as you stated, I get standard deviations of:
Field A, boys: sqrt(.14*500*(1-.14)) = 7.76 students
Field A, girls: sqrt(.06*500*(1-.06)) = 5.31 students
Field B, boys: sqrt(.90*500*(1-.90)) = 6.71 students
Field B, girls: sqrt(.80*500*(1-.80)) = 8.94 students

The maximum possible standard deviation for a 500 student sample is only 11.18 students (using p = .50). I cannot find your number no matter what I try. Have you abandoned the binomial assumption you made in the previous post?
You're using the wrong formula.
To use that one you have to know the mean of the WHOLE population, not the mean of your sampling size, like you used. Unless you mean it as a biased estimator.
The mean of the whole population is what we are trying to estimate. How well does your sampling size represent the whole? That's the error you're looking for.
Also, that's not how you propagate the uncertainty through the calculations.
Sojobo wrote:

Anyway, these numbers are all statistically significant, as I said they would be.
If you use 1000 instead of 100, multiply all the numbers I used by 10, use the US Engineering grad student as the population whole (and use the right formula for the errors) the statistical significance between the genders goes DOWN in my example.

Sojobo wrote:

First, this is just more wankery, and I should probably just ignore it, but I am not going to, though, because it's again so acutely hypocritical. You are the one who has been making up wild numbers that might make your points stick. I have, at every opportunity, claimed that the numbers are likely to be steady, exactly as you have agreed is probably the case.
Call me whatever you want, but I'm the only one that brought any numbers at all to the table, and I'm the only one that did any proper math(well, I'm not so sure how accurate that number i used for the total engineering grad-student population was, but I used the right formula). So, call me a wanker if you like(not that you did, but if you like), but don't allege I am the one making up wild claims.

Sojobo wrote:

Of course they weren't trying to show what I am showing. They are just publishing the data they've collected. I am saying that they wouldn't have published that data if they had only polled a ridiculously small and unrepresentative sample. Therefore, I can assume their numbers are fairly robust.
No, because the poll was not meant to be scientific. You can tell that by the lack of error bars.
Sojobo wrote:
You've told me you agree that the numbers are probably pretty steady. That makes the numbers comparable.
My guessing does not make it true. Wish it did; it would make my life a whole lot easier.
Sojobo wrote:
More numbers would probably make it even more clear that Ember was wrong, but my argument is not with Ember, it is with you. And my contention the whole time is that the numbers given already imply what I say they imply. I have no need for more numbers.
K.
Sojobo wrote:

Go ahead and answer my first question in this post. Tell me that no confidence level implies anything until you reach that magic number of 95%. Perhaps this all comes down to a difference of opinion on what "imply" means. We can sort all this out if you just straightforwardly answer the question.
I did not make the scientific method, I just use it.
If you have a problem with the 95% don't take it up with me, I just work here.

But, more to the point, we don't even know if the 1% mark is even at the 50% confidence level or not.
_________________
A MtG Webcomic
Back to top
View user's profile Send private message
Sojobo



Joined: 12 Jul 2006
Posts: 2443

PostPosted: Sun Sep 16, 2012 10:29 pm    Post subject: Reply with quote

Lich Mong wrote:
I was trying to highlight that it was not 'mine,' but the scientific method's. It was not meant to be an aggressive statement, but just one saying I was not pulling that number out of my own ass.

I was not accusing you of pulling the number out of your ass. I accept that 95% confidence is the effectively universal standard. You are wrong to say it is "scientific method's" standard, though. Scientific method has existed long before man started calculating confidence intervals, much less before 95% became a standard.

Actually, I'm going to go out of order to keep this thread of our conversation coherent:
Lich Mong wrote:
I did not make the scientific method, I just use it.

It is not true that data which yields a 90% interval, but not a 95% interval, is magically "not scientific". It is dishonest for you to pretend I am somehow arguing against the scientific method. It is also dishonest that you twice substituted this claim for answering my question. I asked you if data which yields a 90% interval, but not 95%, is meaningless, and whether it can imply anything. I didn't ask about scientific weight, and I didn't ask about a general standard for confidence intervals.

I'd still like you to answer the questions, by the way, though I don't really expect it. Not answering questions directly seems to be a bit of a pattern for you in this conversation. Plainly, your answer to both of these questions is "yes", and you are just unwilling to say so, because you see it as some sort of concession.

I suppose I cannot entirely fault you, as my claim at the start hasn't been that "the data conclusively and "scientifically" conclude that Ember is wrong", but rather that the data implied that Ember is wrong. So it could be considered a bit of a concession, but you can still argue about whether or not the specific numbers are robust or flimsy, so the debate can continue.

Lich Mong wrote:
Additionally, one sigma is normally 68%, not 90%. But we don't even know if the 1% mark is even better than a 50/50 for that poll.

I'm pretty sure I didn't say 90% is one sigma. Thanks for the correction, though, I guess.

Lich Mong wrote:
We clearly disagree on what 'flimsy' means in this context. What are you using it to mean? That they cared more about Type I errors than Type II? That they cared more about Type II errors then Type I? What? There are lots of ways to 'rate' polls.

I don't think they cared about either Type 1 or Type 2 errors. They were reporting data they'd gathered, and I don't project any motive on them whatsoever (other than my trust that they didn't intentionally publish uselessly weak data).

I care about both Types of errors. I am using flimsy to mean subject to both. I do not have a predefined proportion in mind, though.

Lich Mong wrote:
Its likely they polled more than 100 people, yes. But, what matters is they're dealing with a larger overall population. Lots of people go to gradschool each year.

Both things matter. I think the raw number of people polled (if this is a poll, and assuming standard controls to randomly sample) matters more.

Lich Mong wrote:
I'll bet you my house against yours they did not. You don't normally use a high statistical power for those kind of polls. They're for news papers, not journals.

I wouldn't make that bet. It seems rather irresponsible, as I don't know anything about the Council of Graduate Schools, except that they have a very official-sounding name. As I have said quite a few times, though, I don't think this is a poll. Even if it were, I don't think it would be a "newspaper" poll.

Lich Mong wrote:
Let me restate. If you want to say a larger % of girls dropped out of Engineering gradschool than guys, then what happens in the other fields doesn't matter.
That was your original claim.

I want to say that the difference between the success of men and women in Engineering is not lower than it is in other fields. It is interesting that you have gone back to my original claim even after you brought it up already, and I straightforwardly admitted that I didn't state my whole position originally, and so amended my claim. It is also interesting that you have responded to my amended claim several times, so there is no doubt whatsoever that you know what my contention is. Why have you now returned to discussing the original claim?

Lich Mong wrote:
You're using the wrong formula.

So what is the right formula? I asked you for your calculation last post, but you seem to have overlooked it. I'm still interested. Please post it.

Lich Mong, bolding min wrote:
If you use 1000 instead of 100, multiply all the numbers I used by 10, use the US Engineering grad student as the population whole (and use the right formula for the errors) the statistical significance between the genders goes DOWN in my example.

Use the whatnow? You decided it was fair to change the population whole and not bother mentioning it? If you've bothered to look up the actual population whole, then you should probably go ahead and use the actual statistics that go with it, rather than mixing the number into your example arbitrarily because it gives you the number you want.

Lich Mong wrote:
and I'm the only one that did any proper math

So let me see it. This isn't an attack; I am genuinely curious. I want to know what calculations you've made.

Lich Mong wrote:
My guessing does not make it true. Wish it did; it would make my life a whole lot easier.

You're misunderstanding this argument. I am not saying that wishing makes it true. I am saying:
1) Steady numbers would demonstrate that Ember was wrong, and
2) The numbers are probably steady, therefore
3) Ember was probably wrong.
Or, restated, the numbers imply that Ember is wrong.
_________________
"To love deeply in one direction makes us more loving in all others."
- Anne-Sophie Swetchine
Back to top
View user's profile Send private message
Lich Mong



Joined: 31 May 2012
Posts: 475

PostPosted: Mon Sep 17, 2012 12:27 am    Post subject: Reply with quote

Sojobo wrote:
I was not accusing you of pulling the number out of your ass. I accept that 95% confidence is the effectively universal standard. You are wrong to say it is "scientific method's" standard, though. Scientific method has existed long before man started calculating confidence intervals, much less before 95% became a standard.
Is the percentage for the scientific method used for hypothesis testing in most fields (not all).

And, honestly, saying like you are makes me think you really don't understand what "the scientific method" is. Or, if you find my way of saying that too insulting, that my understanding of it is drastically different from yours.

Sojobo wrote:

It is not true that data which yields a 90% interval, but not a 95% interval, is magically "not scientific". It is dishonest for you to pretend I am somehow arguing against the scientific method. It is also dishonest that you twice substituted this claim for answering my question. I asked you if data which yields a 90% interval, but not 95%, is meaningless, and whether it can imply anything. I didn't ask about scientific weight, and I didn't ask about a general standard for confidence intervals.
No data is "meaningless," and it is dishonest for you to claim I stated any of it was.

However, that data does not have the meaning you are trying to give it. Thus, I called your claims scientifically meaningless. Since, you know, they are.
And, as we have already outlined, they are meaningless for many different reasons, only one of which being the confidence intervals we are discussing.

Sojobo wrote:

I'd still like you to answer the questions, by the way, though I don't really expect it. Not answering questions directly seems to be a bit of a pattern for you in this conversation. Plainly, your answer to both of these questions is "yes", and you are just unwilling to say so, because you see it as some sort of concession.
I am answering your questions, you just don't like my answers because they don't fit into your strawman interpretation of my argument. Your question about weither or not I thought data which met the 90% but not the 95% is meanless or not is a ridiculously loaded question because no data is meaningless. I never even implied any data was. I was only talking about your claims.

So,
I'm doing my best to answer your questions, but it would be easier if you asked real ones instead of loaded ones designed explicitly to try and strawman my argument.

Sojobo wrote:

I suppose I cannot entirely fault you, as my claim at the start hasn't been that "the data conclusively and "scientifically" conclude that Ember is wrong", but rather that the data implied that Ember is wrong. So it could be considered a bit of a concession, but you can still argue about whether or not the specific numbers are robust or flimsy, so the debate can continue.
Except you still can't compare the intake numbers to the outtake ones. So, no.

Sojobo wrote:

I'm pretty sure I didn't say 90% is one sigma. Thanks for the correction, though, I guess.
You did not say it, but it seemed to me you were implying it. I just wanted to make sure we where on the same page.

Sojobo wrote:

I don't think they cared about either Type 1 or Type 2 errors. They were reporting data they'd gathered, and I don't project any motive on them whatsoever (other than my trust that they didn't intentionally publish uselessly weak data).

I care about both Types of errors. I am using flimsy to mean subject to both. I do not have a predefined proportion in mind, though.
Ok, then "no, I think mine aren't more 'flimsy.'"

Sojobo wrote:

Both things matter. I think the raw number of people polled (if this is a poll, and assuming standard controls to randomly sample) matters more.
The ratio matters, so both are equally 'important.'

Sojobo wrote:

I wouldn't make that bet. It seems rather irresponsible, as I don't know anything about the Council of Graduate Schools, except that they have a very official-sounding name. As I have said quite a few times, though, I don't think this is a poll. Even if it were, I don't think it would be a "newspaper" poll.
I see you've not checked your sources. Normally a bad idea. Especially when the guy your arguing against knows he picked them up from someone's blog.

Sojobo wrote:

I want to say that the difference between the success of men and women in Engineering is not lower than it is in other fields. It is interesting that you have gone back to my original claim even after you brought it up already, and I straightforwardly admitted that I didn't state my whole position originally, and so amended my claim. It is also interesting that you have responded to my amended claim several times, so there is no doubt whatsoever that you know what my contention is. Why have you now returned to discussing the original claim?
Or--maybe--you've changed you claims and what your arguing for and against so many times in this conversation I'm just all turned around.

I confuse ease.

Sojobo wrote:

So what is the right formula? I asked you for your calculation last post, but you seem to have overlooked it. I'm still interested. Please post it.
I told you what I used.

I would have linked the wiki page then, but you reacted strongly to my last one. Traditionally, I would have been linking you lots of other wiki pages, but I have been consciously stopping myself as to not offend you.
Again, I used the "standard error."

Sojobo wrote:

Use the whatnow? You decided it was fair to change the population whole and not bother mentioning it?
Sorry, I thought we were still talking about "difference between the success of men and women in Engineering."
Which subject are we talking about? Should I have used the biology majors? All countries everywhere? What are you objecting too?
Sojobo wrote:
If you've bothered to look up the actual population whole, then you should probably go ahead and use the actual statistics that go with it, rather than mixing the number into your example arbitrarily because it gives you the number you want.
I guestmatted the total US Engineering gradstudent population based on some quick research I did. Hence the term "back of the envelope," and my disclaimer of the accuracy of my number.
Also, I don't know how many people were polled the poll in question, so I can't do what you're asking me to even if I did have the real number, which I don't.
Sojobo wrote:

So let me see it. This isn't an attack; I am genuinely curious. I want to know what calculations you've made.
I don't have the # I used for the grad population, and I don't feel like rederiving it for you.

What I have already told you I feel is sufficient. If you feel otherwise you should get the real number yourself and prove me wrong.

Sojobo wrote:

You're misunderstanding this argument. I am not saying that wishing makes it true. I am saying:
1) Steady numbers would demonstrate that Ember was wrong, and
2) The numbers are probably steady, therefore
3) Ember was probably wrong.
Or, restated, the numbers imply that Ember is wrong.

"Steady" and "don't vary year to year by more than 1%" is not the same thing.
I, in fact, find it highly likely they change more then that each year, but I would still call them 'steady.'
_________________
A MtG Webcomic


Last edited by Lich Mong on Mon Sep 17, 2012 2:08 pm; edited 1 time in total
Back to top
View user's profile Send private message
Rothide



Joined: 14 Jul 2012
Posts: 852

PostPosted: Mon Sep 17, 2012 3:48 am    Post subject: Reply with quote

THREAD POLICE! We have had a report of a thread about to break 25 Pages, we are getting noise complaints from the lurkers and roughhoused words on the posts. I want this here thread cleared out in 10 minutes!
Back to top
View user's profile Send private message
Darqcyde



Joined: 11 Jul 2006
Posts: 10566
Location: A false vacuum abiding in ignorance.

PostPosted: Mon Sep 17, 2012 3:50 am    Post subject: Reply with quote

OH SHIT! IT'S THE FUZZ!!!!

*RUNS FOR COVER*
_________________
...if a single leaf holds the eye, it will be as if the remaining leaves were not there.
http://about.me/omardrake
Back to top
View user's profile Send private message Send e-mail Visit poster's website AIM Address Yahoo Messenger
Heretical Rants



Joined: 21 Jul 2009
Posts: 5344
Location: No.

PostPosted: Mon Sep 17, 2012 3:50 am    Post subject: Reply with quote

I honestly have no idea what you guys are arguing about. Neither of you think that the numbers support what Ember was saying, am I right?
_________________
butts
Back to top
View user's profile Send private message
Lich Mong



Joined: 31 May 2012
Posts: 475

PostPosted: Mon Sep 17, 2012 4:11 am    Post subject: Reply with quote

Ember778 wrote:

I believe they're sufficient to explain things like gender roles. Take my field of engineering for example. Engineering is a male dominated field. This is not because males are just employed more, but because more males pick engineering as a major than females.

Now this isn't because males are in any way smarter than females it's just males are normally better at math and sciences, and also women are normally less ambitious than males and engineering is pretty hard and requires a lot of effort.

Ambitiousness can be directly tied to testosterone, and males are just generally better at math and science because of how our brains work (psychology people could explain it better than me).
Lich Mong wrote:

http://2.bp.blogspot.com/-3GsUJIOjgXk/Tnuq3OOcGGI/AAAAAAAAPvg/LqfA2CsO7Qs/s1600/grad2.jpg
http://1.bp.blogspot.com/-zgTqUtZFb7c/Tnuq2pgF5PI/AAAAAAAAPvc/hPKWMOMWLgc/s1600/grad1.jpg
Hu, with all of the women in my year I would have not guessed those numbers....
Anyway.
Your logic, Ember778, seems to indicate that about the only things men are good at are math, science, and engineering. And that women are--on average--smarter than men.

Is that what you're trying to say?
Ember778 wrote:
Well I wouldn't put it like that exactly, but in essence yeah.

I mean I wouldn't have said that the only thing men are good at is math, science and engineering, but it's primarily what they're good at.

Also yeah I've found that women on average have higher intelligence than men.

That doesn't mean there are some really really dumb girls out there just that from what I've seen women tend to be a bit brighter in non-math/science stuff.
Lich Mong wrote:
Well, I was expecting a special pleading argument out of you, but instead you gave me an internally consistent bullet bite. So, I guess I'll have to at least tip my hat to you for exceeding expectations.
Sojobo wrote:
Ember778 wrote:
Now this isn't because males are in any way smarter than females it's just males are normally better at math and sciences, and also women are normally less ambitious than males and engineering is pretty hard and requires a lot of effort.

If your claim were true, if men were better at math and sciences, and specifically at engineering (because it is so hard), then we would expect to see that women, enrolling in graduate school for engineering, would be less successful than men doing the same thing.

Lich Mong just quoted some statistics at you. You responded to that post, so I'm going to assume you actually read them (possibly a risky assumption on my part).

Those statistics say that women make up 22.3% of graduate enrollment for engineering, and 23.2% of the doctoral degrees. This is very different than we guessed, based on your claim being true. It should actually be quite an unlikely result, given the large number of trials. Your claim probably doesn't have very much merit.
Lich Mong wrote:
Guys, hey guys, guys, I've got something to say guys, guys....

...I hate to be the one to break it to everyone, but there is nothing reasonable about this line of reasoning--like--any of it (except maybe that link of Halen's).

First, the number of people graduating is not equivalent to the number of people being admitted; to make the kind of inference you're making, Sojobo, you would need to integrate these numbers over time, not look at the instantaneous single year flux.
But, more to the point, they don't even have error bars on those statistics. We don't know how many sigma 1.0% is. So, we can't even say if less dropped out even if we made the assumption we can't make that the number entering is equal to the number leaving.

Ember778's reasoning is equally groundless in that these numbers do not take into account societal factors like gender cronyism, nurturing differences, and field prejudices. All they could be saying is that engineers are more sexist and turn more female students off to the subject before they even think about applying to grad-school, and then reject more when they do. Or they could be saying that grad-schools are under pressure to admit, and pass, more females.
Additionally, Ember778 has already stated--pretty clearly--that his reasoning is based on anecdotal evidence. When you make declaimers like "from what I've seen" it means you're just speaking from personal experience.
Like, from what I've seen the best engineers are closed-minded hyper-conservative male-chauvinists, but that does not mean much of anything because it's simply my personal experience, not a rigorous statistical statement.

Anyway, I gave Ember778 passing marks for having his reasoning be internally consistent(which in this day and age is fairly impressive for an opinion; good job), but it's certainly not based on anything reasonable.
Sojobo wrote:
Lich Mong wrote:
First, the number of people graduating is not equivalent to the number of people being admitted;

Nor do I think it is. Nor do I imply it is. I don't know why you're correcting me about this.

Lich Mong wrote:
to make the kind of inference you're making, Sojobo, you would need to integrate these numbers over time, not look at the instantaneous single year flux.

This is not true. Several years of data (especially tracking one class' enrollment through that class' graduation) would give much stronger evidence, but that doesn't mean a single year is meaningless.

Lich Mong wrote:
But, more to the point, they don't even have error bars on those statistics. We don't know how many sigma 1.0% is.

Again, you're looking for info that would give much stronger evidence, or rather much more precise evidence, but this doesn't preclude the numbers, as given, strongly implying that Ember is wrong.

Then it all when to hell because I couldn't let it just pass:
Lich Mong wrote:
Sojobo......... sigh......
Based on what I just spent the last 4 days doing, I really don't want to go into this too deeply here because I likely will end up hurting myself. But, sufficient to say, without knowing the standard deviation on those statistics you absolutely cannot say anything meaningful about the difference between 22.2% and 23.2%

Additionally, your other statements simply aren't reasonable unless you can make the assumptions I outlined (which you can't). We simply don't know if 1% of the people coming 4-5 years before were female or if 99% were female. The numbers were probably close to 20%, but without knowing (and knowing the error) we can't even begin to make statements like you made and are making. Did more girls enroll in 2010 than years before or did less? How many more or less? IDK.

I mean "given the large number of trials" & "strongly implying" com'on now.... I don't even know were those numbers came from, I just did a google image search because I wanted to see what kind of merit Ember778's statements about enrollment had (at the time I honestly thought he was wrong about it).


And, from there it was a downward spiral....
_________________
A MtG Webcomic


Last edited by Lich Mong on Mon Sep 17, 2012 4:13 am; edited 1 time in total
Back to top
View user's profile Send private message
mouse



Joined: 10 Jul 2006
Posts: 17282
Location: under the bed

PostPosted: Mon Sep 17, 2012 9:49 pm    Post subject: Reply with quote

i hate to ruin a good argument with actual facts, but it turns out that the council of graduate schools has an actual website. with links to their reports and methods and all like that.



spoiler alert! they don't poll graduate students at all.
_________________
aka: neverscared!
Back to top
View user's profile Send private message
Display posts from previous:   
Post new topic   Reply to topic    Sinfest Forum Index -> Sinfest All times are GMT
Goto page Previous  1, 2, 3 ... , 23, 24, 25  Next
Page 24 of 25

 
Jump to:  
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum


Powered by phpBB © 2001, 2005 phpBB Group