Math Survey Question

[Zetta_x]

Thanks Sterling for clearing it out!

Ahh, it doesn't have to be so Sake I can find four observations (0, 20, 0, 20) that has a large enough variability to still include 20 as not an outlier but the average is 10 =D.

Tom, if you see the statistics homework I get in my Ph.D. program, you will understand that this problem was more for research ideas than homework help.

---

I want my research to be able to statistically model someone's decision factors and be able to predict their next action. Everyone has a set of logic and problem solving skills that ties in closely to their decision making. Maybe I'll share my research ideas, but this topic was to brainstorm the type of logic people use.

--

Edit: I don't even know how to accurately explain the classes I am taking.

 

[Sterling]

When I saw Saken's post, I thought to myself that 20 is an acceptable margin in this grouping. 5 observations didn't really seem like a large enough sample size to me to count that 20 as an outlier figure. Just to be clear though, I've never taken any classes on statistics, but I understand a few terms here and there.

 

[Zetta_x]

If you ever need any advice on statistics, let me know! But you are right, with a sample size so small, outliers are hard to identify. It's easy if you have like 100,000 independently identically distributed observations (by the central limit theorem, the mean of these 100,000 observations would be normally distributed). Then, if you have something about 5 standard deviations away from the mean with these observations it's so easy to see it as an outlier (or something that was not from the original population).

On the other hand, think of 2 points. If you have 1 point that says 50 and another point that says 300; how do you know if any are outliers. If you guess one is an outlier, why is 50 any less of an outlier than 300 is? Both can't be outliers now can we really identify many properties of where they come from.

 

[Sterling]

I mean if that 20 was actually a 100, then I might consider that an outlier. The acceptable margin would look to me to be passed. Since I can't tell what we're observing, I'd think that this 20 would be acceptable given the majority of the others are just half the figure and not a fourth or a fifth of it.

 

[Coto]

On the other hand, think of 2 points. If you have 1 point that says 50 and another point that says 300; how do you know if any are outliers. If you guess one is an outlier, why is 50 any less of an outlier than 300 is? Both can't be outliers now can we really identify many properties of where they come from.

If I may join the point you're trying to explain i'd say:

Spoiler

1 point being 50
1 point being 300

means:

2 points are made of 350, or:
350/7=50. (factor unit, being 50, 1 out of 7)

(i picked to divise by 7 because that number divided by both points's numbers [300+50=350] gives one factor unit, or matches the 1st outlier value, and the 2nd too.

if outliner 1=[50 1/7]
then outliner 2=[300 6/7]

that is, assuming i have 2 points like you've described. Uhm that's how I got it, it may be completely wrong, but it was fun to give it a thought.

 

[Mantis41]

Spoiler

(4*10+20)/5

 

[Alexrose]

Alex, the purpose is that every vote is independent. If you show your answer, the votes have dependence amongst them. Do you know the definition of what it means for each observation to be uncorrelated?

For university I just found the masses of the W and Z bosons by making cuts Tevatron data based off of Monte Carlo simulations to remove background particles and then fitting a convolution of Gaussian and Relativistic Breit-Wigner distributions to the Z mass and a Landau distribution to the W mass, then found the efficiencies and random errors and removed the systematic errors. Yes I know what correlation is. An 8 year old knows what correlation is.

And what's more, I know that what you meant is "independent", not "uncorrelated". If the vote was biased, you'd expect a great deal more correlation.

Not that there would be any correlation whatsoever in this "experiment". You're essentially measuring two quantities, frequency and "option", and despite the "option"s having numerical values, there's not going to be a continuous distribution. It's just going to be one big definitive peak at the correct value and a couple of much lower peaks for misconstrued values, which numerically have very little to do with each other in the sense that they won't follow some formula you can fit.

There once was a post on facebook with elementary math (the one that had the order of operations). I think as much as 1/3 to 1/2 half got it wrong.

There's one of those every week, but go figure. The world is full of morons who don't know the difference between "its" and "it's", think "weird" is spelt "wierd" and think GCSE maths is really difficult.

Thank you for being a douchebag and not put spoiler tags when you clearly see that I have asked to do so.

Like anyone was going to scroll down the page before they answered the poll.

 

[Coto]

uhm alexrose you were reported. have a nice day, and stop thinking you're too awesome, because you'll rotten and have a shameful life alone.

zetta_x, I lov'd the topic.

 

[Alexrose]

You are taking observations, therefore omit the 20 value as outlier.

You can't just cherry pick what you think is an outlier.

You have to use an analytical method like Chauvenet's criterion.

So, Chauvenet's criterion states that if the probability of the result being at that point on a gaussin distribution is greater than 1/2n where n is the sample size, the result is valid.

So, we know the mean is 10 and the result in question is at 20, so now we just need to determine what standard deviation would call that an outlier.

Now, we need a result where the probability is greater than 0.1, because our data sample is 5, so 1/2n = 0.1.

If you look in this table:

https://statistics.laerd.com/statistical-guides/img/normal/normal-table-large.png

We need to find a result which is 1-0.05 (rather than 0.1 because of the single tail) from the mean to find the standard deviation that would cause this to be an outlier.

In this case, it's 1.64 sigma away from the mean that would give an outlier.

So 10+1.64sigma = 20.

sigma = 6.09.

So as long as the standard deviation is greater than 6.09, the data is valid.

So, for instance, if the measurements were 16, 15, 5, 4, 20, the last point would be perfectly valid.

But since we weren't given a standard deviation, if you're asked to find the mean, it's a fair assumption that it's not anomalous.

uhm alexrose you were reported.

You say that like it means anything. What did you put in the report form, "I don't agree with this guy, ban him plz"?

 

[Coto]

no uhm it was more liek "uhm this guy knows everythung and bashes everyone zinking he'z awwzomeW with his superb amount of biblic suppossed-to-be-true arguments, but in the end they have no sense at all. I mean do we have to read such stuff?

IIRC that's is flamming.

No sense, or you don't understand it?

I think flaming is more like acting so demeaning that you ask an adult if he's aware of what correlation is. I gave a perfectly objective response to someone who'd just called me a douchebag and asked me if I was aware of something a child could tell you, and I'm flaming? Right.

Oh I don't need any gaussian or chauvenet's criteria to resolve easy or complex tasks on real life. If those guys were to know you used them to debate something that's completely irrelevant, yet easy, they'd beat you with a stick, i mean you're smart enough to live with your own perception/logical point of view of how matter/numbers works/ act right, or they must be on some book so you feel better and not sorry about "what you are good at"?.

So no, i don't know about them, and it's irrelevant to show off others how much stuff you know if you'll live alone in that world because you're unable to cooperate with OP (or real life people), you'll realize it someday, trust me.
/offtopic

uhm zetta_x, it would be a good idea you could post such maths/statistics exercices more often

 

[Alexrose]

no uhm it was more liek "uhm this guy knows everythung and bashes everyone zinking he'z awwzomeW with his superb amount of biblic suppossed-to-be-true arguments, but in the end they have no sense at all.

No sense, or you don't understand it?

I think flaming is more like acting so demeaning that you ask an adult if he's aware of what correlation is. I gave a perfectly objective response to someone who'd just called me a douchebag and asked me if I was aware of something a child could tell you, and I'm flaming? Right.

Oh I don't need any gaussian or chauvenet's criteria to resolve easy or complex tasks on real life.

Obviously you don't need statistical analysis for anything to do with real life, but if you were in any field of science and you didn't know how to analyse your data, you would fail your degree.

If I went into a lab interview and said "this data point is an outlier, I can just see that, I don't need chauvenet's criterion to resolve an easy thing like that", they would dock my mark 5%. I'm not advocating using these things for every day life - there's no application for them in day to day life, but if people used your immature attitude towards dealing with problems rather than the scientific method, we would be 200 years behind in technology. If it's a statistical and numerical problem, you use statistical analytic methods.

If those guys were to know you used them to debate something that's completely irrelevant, yet easy, they'd beat you with a stick, i mean you're smart enough to live with your own perception/logical point of view of how matter/numbers works/ act right, or they must be on some book so you feel better and not sorry about "what you are good at"?.

It's funny, because you say it'd be difficult to cooperate with me, but you're unable to even write a sentence that's legible. I don't think anyone would ever be able to cooperate with someone who can't even convey a sentence. I have no idea what you thought you were trying to say here, but you managed to say nothing comprehensible. I particularly enjoyed the part where you made your text bold and highlighted nothing.

 

[GeekyGuy]

Stay on-topic, cut the bullshit.

 

[Tom]

This says otherwise. >.>

That's still not in the topic, thats in the thread.

 

[Sterling]

That's still not in the topic, thats in the thread.

EDIT: I take all this back. Fucking avatar stand-ins are confusing me. Sorry Tom.

 

[Zetta_x]

Alex, it seems you know how to use statistics but I'm not sure if you just use statistics or you fundamentally understand when it is appropriate to use it?

If something is independent, then it has no correlation. Not necessarily the other way around obviously... I grew up in America and a 20 year old does not know what correlation is yet alone anyone younger. You would be surprised about how many people get this question wrong. While the spoiler tag would prevent from someone accidentally seeing the answer, I'm guessing the spoiler tag did not stop most people but it would prevent someone who scrolls down real quick but decided to vote anyways without looking at the answer. It's like a confirmation screen if you want the answer.

 

[Alexrose]

If something is independent, then it has no correlation.

I never suggested otherwise.

You were implying that people reading my post would contaminate your data. In other words, they would read my post and their answer would become dependent on my answer.

You said "Do you know the definition of what it means for each observation to be uncorrelated?"

Well that has absolutely no relevance here. There is zero correlation. You've provided a list of arbitrary choices. It's like saying to a test sample "Do you prefer cake, biscuits or sweets?". You can't correlate that to anything; you only have one variable: frequency, and despite your given choices being numerical options, they are completely arbitrary. There's no case of the central limit theorem stacking results to the centre of the data or anything like that; all data is entirely dependent on someone either making a calculation correctly or making a calculation correctly, which isn't any more probabilistically likely to adopt low values as it is to adopt high values.

If people reading my post contaminated your data, it didn't change how much it "correlates" at all, because there is no metric you can compare it to, unless you plot "numerical value of choice" (to which you don't have results on a continuous scale) against "frequency of choice", which will produce absolutely meaningless data. It would change whether or not the results were independent, and therefore some ratio, not correlation.

 

[Tom]

EDIT: I take all this back. Fucking avatar stand-ins are confusing me. Sorry Tom.

 

[Zetta_x]

I never suggested otherwise.

You were implying that people reading my post would contaminate your data. In other words, they would read my post and their answer would become dependent on my answer.

You said "Do you know the definition of what it means for each observation to be uncorrelated?"

Well that has absolutely no relevance here. There is zero correlation. You've provided a list of arbitrary choices. It's like saying to a test sample "Do you prefer cake, biscuits or sweets?". You can't correlate that to anything; you only have one variable: frequency, and despite your given choices being numerical options, they are completely arbitrary. There's no case of the central limit theorem stacking results to the centre of the data or anything like that; all data is entirely dependent on someone either making a calculation correctly or making a calculation correctly, which isn't any more probabilistically likely to adopt low values as it is to adopt high values.

If people reading my post contaminated your data, it didn't change how much it "correlates" at all, because there is no metric you can compare it to, unless you plot "numerical value of choice" (to which you don't have results on a continuous scale) against "frequency of choice", which will produce absolutely meaningless data. It would change whether or not the results were independent, and therefore some ratio, not correlation.

I envy your ability to be wrong and actually have more bull shit to support it.

---

The list is in no way arbitrary. 10 is the answer you would find when you think it's a trick question. 12 is the correct answer assuming your logic is sounds. 15 is the correct answer when you have unsound logic. The rest of the answers form one group intended to catch people who randomly guess it or put in an answer with absolutely no thought.

Correlation can be talked about. Does someone else's vote have any influence on another person's vote? That's correlation; What I was talking about, is seeing the correct answer someone is more influenced to guess the correct answer. There are correct answers and there are not correct answers, there is nothing arbitrary about it.. In other words, this can be mapped out in a time domain. What are you talking about no metric... Are you suggesting the central limit theorem does not hold for a binomial distribution (or a sum of bernoulli distributions)? Because it's just frequency right?

Jesus... I'm done.

You have proven to me that you know how to use statistics but you lack in the most basic foundations; you don't have a clue what assumptions you are making when using your statistics, you just know what people tell you.

 

[Coto]

That's what I meant. Anyway, I hope you do another thread like this zetta, i've really enjoyed it.

 

[Zetta_x]

That's what I meant. Anyway, I hope you do another thread like this zetta, i've really enjoyed it.

This thread turned out interesting! I got what I wanted and more! What I have always liked about GBAtemp is that people are not afraid to show themselves. In life, people are fake. It bothers me so much, but people tend to hide a lot in social situations. On GBAtemp, we show our true selves and that type of data is much more valuable.

What kind of thread would you like Coto, a math one in general (this one was more about logic than anything), something that is difficult?