can someone help me with this problem? maths

Post Reply New Topic

I am doing an online quiz about "which country fits you best". There are questions that determine your scores (in percentage form) for each variable and then I need a comparison of your score and the country for each variable. each country is like this (example):

china F=8,5 H=-2 T=1,2 A=6 B=7,2
nigeria F=3 H=9 T=-0,5 A=2 B=6
there are 30 options of countries

and you get a score like this in the first part:
you F=28% H=3% T=15% A=72% B=83%

I had intended to do this quiz another way but I couldn't find the format I wanted. If anyone knows a great free site for creating quizes let me know!

okay, so in the second part of the test there are questions like: 'what was your F score?' then you can choose 4 different answers (I don't know yet where to 'cut' the scores in those answers). Each answer I can select countries and give points from -10 to +10 (I'll call this country score). so that the country with more points is the result in the end.
My doubt is:
- what would be the best cuts in the answers of the second part? how do I calculate this?
- how can I adjust the country score so that the results are as accurate as possible? (meaning the result is the country that has variable numbers the closest to your numbers?)

so I need help with this! The quiz is only meant for fun but I got interested in how to solve this. problem is I know nothing of statistics. Also if you didn't understand my explanation please ask.

Without knowing what the variables mean, the question is less than useless.

You're right. The answer is 3. It is always 3.

When I was an undergraduate in Math in the mid 1970s, I fell asleep one day in class. The prof got mad and asked me a question to wake me up. I started to reply "3" but figured that was unlikely to be the answer. It turned out that the answer was "3".

From now on, the answer is "3".

So "3" is the answer.

So, eric is being facetious because your answer makes no sense given your question.

You really need to learn statistics to answer your questions. I think this whole thing is your wake up call from life. You were meant to study statistics and you should pursue it. Then you will have the answer to your questions and understand eric's response.

And it will be fun if you keep this spirit of investigative interest going.

Good luck!

Edited to add:
It's a bit like you having asked "How long is a typical piece of string?"
eric replied "what kind of string? used for what?"
And you replied "it's a string the ties things together--what difference can that make?"

So "3" is a perfectly good answer. 3 inches, 3 meters, 3 kilometers. All good.

There is something that may or may not be of interest regarding machine learning.

Suppose that you have a data set made up of a great many individual factors that may or may not correlate to the result or outcome you want.

Suppose that you have a large dataset containing a variety of things that may or may not relate to the odds of developing lung cancer and you want to be able to use that large dataset to determine the odds of developing lung cancer for a new case.

The data may include things like:
number of cigarettes smoked daily
current age
current weight
age began smoking
works in underground mines (yes or no)
number of cigarettes spouse smokes daily
heats house with wood fireplace (yes or no)
severity of allergies
severity of sinus infections
platelet count
blood serum vitamin d count
stored iron levels
cholesterol count
...

And for the sample, you would have whether or not they had lung cancer.

One approach would be to build a neural net within each input that creates a number representing either lung cancer or no lung cancer and the train that on the data set.

Then using the same data for new cases, you can run them through the neural net and have it provide you with a result indicating whether or not they are likely to develop lung cancer.

You could do the same thing based on political views and background to predict who is more likely to vote in an election.

Of course, neural nets are not all-knowing magical devices -- they are relatively simple classifiers and the results may or may not be of use.

So if these different variables are for things like:
1) likes the beach
2) parties heavily
3) likes museums
4) speaks Spanish
5) has savings of at least $100,000
6) number in family
7) enjoys theater
enjoys classical music
9) enjoys winter sports
and so on, and you are trying to find a country that fits them best, then one way would be to try to identify a large number of variables and then get a large number of people of like those who will be taking the test to answer questionaires about those variables. Then create a neural network for each country and train each neural network on that set of data. You could then use the questionaires from others as input to the neural networks to find out whether or not they would match others who went to those countries.

Keep in mind that you would likely not really be figuring out which country fits them best but which country people who answered the questionaires in the original data set thought fit them best. That is, instead of finding that Norway fit the test taker best, you would find that people with similar interests thought that Norway fit themselves best.

Of course, for a simple on-line test of some kind, this is overkill. Instead of spending time and effort trying to find the best way to apply the data to the test takers, just come up with something off of the top of your head and use it.

By the way. Percentage is a fraction with the denominator = 100.