help with best fit line

Reply #19

notmyday — Mon, 29 Jan 2018 21:33:12 GMT

Glad you got some rest hopefully you are back on track. Hope them bricks didnt hurt. stupid part you are not even close. extreme tired you was. Your programming can wait its not going anywhere while you take a break and get some rest.

Reply #18

RL-RANDOMLOGIC — Mon, 29 Jan 2018 14:30:25 GMT

nmd

I found the problem, when the scatter plot is printed the values in the data file are scaled to

the graph. When the regression was calculated it was still using those same values. What I

had to do was swap the original data values with the newly calculated Y values of the dots

on the scatter plot. They have to be inverted as the higher a dot falls on the chart the lower

the y value is. Still working it out but I don't see a problem with it now. Funny thing is I woke

up this... [ More ]

Reply #17

notmyday — Mon, 29 Jan 2018 10:37:00 GMT

Now your getting somewhere. Yah

Reply #16

RL-RANDOMLOGIC — Mon, 29 Jan 2018 06:57:56 GMT

Works perfect with a different data set. I had to rescale the dots on the scatter plot which are not exact but

got the same best fit line and all the correct values as a hand workout. I know what the problem is, how to

work around it may be a little tougher.

RL

The 10 data points used in this example are included below pic.

10 sample example.

1,14

6,9

9,4

12,1

4,12

5,8

6,6

11,3

2,10

8,3

Reply #15

RL-RANDOMLOGIC — Mon, 29 Jan 2018 02:13:47 GMT

When using input with comma delimited data it does not matter. I got to step back and think about

this some more. It's weird to say the least.

RL

Reply #14

notmyday — Mon, 29 Jan 2018 02:10:55 GMT

Did you catch the data part i posted.

Oh there was one thing i found in your data when i copied it the last line i had to put a space at the very end of it because each of your lines had an extra space except for the last line.

Reply #13

RL-RANDOMLOGIC — Mon, 29 Jan 2018 02:09:16 GMT

I tried 3 different versions and all 3 produce the same exact errors in the calculations.

Even when I copy and paste the code I posted with the string name value$ fixed. It

appears that the compiler is producing the problem when ran on this computer. I will

try a different PC and see what happens.

RL

Reply #12

notmyday — Mon, 29 Jan 2018 02:02:51 GMT

It is a 7.29 mb file i downloaded it on 9-24-17 qb64 x32

Reply #11

RL-RANDOMLOGIC — Mon, 29 Jan 2018 02:01:08 GMT

I think it's a variable conflict, just got to find it. This code was pulled out of the main program

which is around 10K lines. Which version of QB64 are you using? The one I am using for this

project is a dirty build which may be causing the problem.

RL

Reply #10

notmyday — Mon, 29 Jan 2018 01:59:10 GMT

Oh there was one thing i found in your data when i copied it the last line i had to put a space at the very end of it because each of your lines had an extra space except for the last line.

Reply #9

notmyday — Mon, 29 Jan 2018 01:54:49 GMT

I copied the code in notepad first then put it in qb64 but i changed the spelling in notepad first then saved as reg.text and your data as REG1.txt

Reply #8

RL-RANDOMLOGIC — Mon, 29 Jan 2018 01:50:41 GMT

Looks like some sort of glitch when I copied and pasted the code. Still having the same problems

when I run it. Anyway, it does give me a starting point to.

RL

Reply #7

notmyday — Mon, 29 Jan 2018 01:44:35 GMT

I have caught myself doing the same thing only to fine i missed a key when typing. Get some rest it is always good to step away and look with a fresh set of eyes later after the much needed rest.

Reply #6

RL-RANDOMLOGIC — Mon, 29 Jan 2018 01:41:36 GMT

P.S.

I never noticed that, good catch. Maybe if I ever get caught up on my sleep I will see these things.

RL

Reply #5

RL-RANDOMLOGIC — Mon, 29 Jan 2018 01:37:35 GMT

The fact that your run produced different results gives me an idea where to start looking.

RL

Reply #4

notmyday — Mon, 29 Jan 2018 01:31:26 GMT

I tried studying your code and the spelling of the value was all i could see hope that helped.

Reply #3

RL-RANDOMLOGIC — Mon, 29 Jan 2018 01:29:46 GMT

nmd

Interesting. It's got me thinking it's a variable problem, the results you get look correct. I will use

your results and see what it looks like when I add the best fit line. The correlation coefficient R

looks like it matches what I see in the scatter plot.

RL

Reply #2

notmyday — Mon, 29 Jan 2018 01:22:59 GMT

I am using your data but getting different results. In your code at the bottom not sure if it is suppose to be that way but i changed the spelling in the word value in all positions and my outcome is different than yours. see below.

Reply #1

RL-RANDOMLOGIC — Mon, 29 Jan 2018 00:12:38 GMT

Source code

Sorry for the block format, whenever I post code it comes out this way.

This code sample generates the following

Mean of x values- meanx

Mean of y values- meany

SD of x values- sdx

SD of y values- sdy

y - m * x + b,

m - m

b - b

the correlation coefficient R = R

Note! does not generate the scatter plot.

To run copy and paste the data from the first post into notepad and save

it as REG1.TXT or create your own data set. This is coded is QB64.

S... [ More ]

help with best fit line

RL-RANDOMLOGIC — Sun, 28 Jan 2018 15:49:44 GMT

Hi all

Need some help with best fit line for a regression tool. The vertical graph is the value and the horizontal is the

game/drawing. The graph is 4x4 pixel and the data used is shown below the picture. First I am thinking the

problem may be with the data as both x and y are percentages and the graph uses whole units for the Horz

and percents for the vertical. Anyway all the values are shown at the top but I can't fit M an B. Maybe I just

need some sleep but if anyone has some s... [ More ]