meta user
network profile
| bio | website | |
|---|---|---|
| location | ||
| age | ||
| visits | member for | 1 year, 5 months |
| seen | May 22 at 14:57 | |
| stats | profile views | 56 |
|
May 13 |
asked | Getting the degree of overlap between each pair of classes |
|
Apr 30 |
comment |
Line of best fit (Linear regression) over vertical line If Y varies and X varies by very slightly, I'll also get the same problem to estimate that line or not ? |
|
Apr 30 |
comment |
Line of best fit (Linear regression) over vertical line Or (1) is it possible to just flip/swap X and Y when the coefficient of correlation is close to 1 ?! (2) It will be close to 1 if we have the case that I shown on the image that I posted ? |
|
Apr 30 |
comment |
Line of best fit (Linear regression) over vertical line Ok, I'll test using the coefficient of correlation and add some random noise to x (is this want you call jitter?) is the correlation is close to 1. I'm using python, sklearn, and the format of the data is just a matrix. |
|
Apr 30 |
comment |
Line of best fit (Linear regression) over vertical line I can not catch this special case where Y do not depends on X, can I ? How ? Or should I always add some noise to my X values ?! |
|
Apr 30 |
comment |
Line of best fit (Linear regression) over vertical line Basically, I want to draw a line which is as close as possible to all points (as a linear regression will do in case where Y depends on X) even if there is no correlation between Y and X values ... I want a method which I can apply whatever is the correlation of Y and X (even there is or there no correlation between them) ! How to do that ? |
|
Apr 30 |
comment |
Line of best fit (Linear regression) over vertical line And this is not for prediction, it is for a pattern recognition problem, where my points are some pixels defined by their coordinates x,y on the image ... |
|
Apr 30 |
comment |
Line of best fit (Linear regression) over vertical line This is just a special case, we can have data where X do not change and Y values changes, or not. I want to draw a line of best fit in all cases ... whatever is my data. |
|
Apr 30 |
comment |
Line of best fit (Linear regression) over vertical line If I give you some data points where X do not change while Y values change. Y do not depends on X. I'm not looking to learn a predictive model. However, I can draw a vertical line which passes over all that points. This is just what I want ! |
|
Apr 30 |
comment |
Line of best fit (Linear regression) over vertical line So to be more clear: My objective is not to predict some y values based on new values of x, my objective is rather just to get a line which passes as close as possible to all points that we have (even if Y are independent of X ...). |
|
Apr 30 |
revised |
Line of best fit (Linear regression) over vertical line added 614 characters in body |
|
Apr 30 |
comment |
Line of best fit (Linear regression) over vertical line I think that I was not very clear in my question. I want to fit the data whatever is the correlation between X and Y (this is not regression), the objective is not to predict some y values based on new values of x, my objective is rather to get a line which passes as close as possible to all points that we have. |
|
Apr 30 |
revised |
Line of best fit (Linear regression) over vertical line added 614 characters in body |
|
Apr 30 |
comment |
Line of best fit (Linear regression) over vertical line I want to estimate line of best fit for another purpose, not to evaluate how much X and Y are correlated (it is more for a pattern recognition problem, where my data-points are some pixels defined by their coordinates x,y on the image). So I want to fit the line even if X values are not correlated to Y values. And if I just always swap X and Y to be able to fit the line, the next time if I have points lying on a horizontal line (Y = [3,3,3,3], X=[3, 6, 23, 30] for example) the if they are swaped I'll get again the same problem. I want to always fit the line whatever is the correlation of X,Y |
|
Apr 30 |
revised |
Line of best fit (Linear regression) over vertical line added 122 characters in body |
|
Apr 30 |
asked | Line of best fit (Linear regression) over vertical line |
|
Apr 24 |
accepted | Transforming the distance value from a center, to a probability value |
|
Apr 23 |
comment |
Transforming the distance value from a center, to a probability value can you give me an example according to my problem which show where $p_{x,c1}$ would be considered as a significant real probability, and where it would be considered as just a value in [0,1] ? So that I can understand better what you mean. |
|
Apr 23 |
comment |
Transforming the distance value from a center, to a probability value If I estimate the distribution of the distances (previously observed) using some method (I don't know which), is it possible to have the cdf which will reflect the probability I'm looking for ? |
|
Apr 23 |
comment |
Transforming the distance value from a center, to a probability value @whuber Because I'll do inside a loop, something like: if( uniform_random([0,1]) < $P_{x,c1}$ ) then "create a a new center at x"; else "e.g. assign x to c1" |
