Curve fitting - Mitscherlich model

M

mjbeam

Well-known Member

Joined

Apr 6, 2002

Messages

756

• Jun 25, 2017

• #2

Put the X values in column A, the Y values in column B.

Solve the equation log(A-Y) = log(A) - cX for A and put this equation in column C.

Plot the data using a scatter plot.

 

Upvote 0

P

phmalu

Board Regular

Joined

Jun 21, 2017

Messages

52

Office Version

1. 2019

Platform

1. Windows

• Jun 25, 2017

• #3

mjbeam said:

Solve the equation log(A-Y) = log(A) - cX for A and put this equation in column C.

Click to expand...

No idea how to do that
Here I have some examples of X and Y:

X	Y
3	31,32
2	33,45
3	31,65
3	31,32
3	25,73
6	30,52
4	35,6
6	33,59
7	33,46
2	32,92
5	37,2
9	37,59
8	42,92
7	39,46
4	21,99
8	25,2

<colgroup><col width="64" span="2" style="width:48pt"> </colgroup><tbody>
</tbody>

 

Upvote 0

M

mjbeam

Well-known Member

Joined

Apr 6, 2002

Messages

756

• Jun 25, 2017

• #4

You can solve your equation online here: https://www.symbolab.com/solver/equ...t(A-Y\right)=\log_{10}\left(A\right)-C\cdot X

This is assuming you are using base 10 for your log function. If it is natural log, use this:

https://www.symbolab.com/solver/equ...-Y\right)=\log_{e^{1}}\left(A\right)-C\cdot X

 

Upvote 0

P

phmalu

Board Regular

Joined

Jun 21, 2017

Messages

52

Office Version

1. 2019

Platform

1. Windows

• Jun 25, 2017

• #5

mjbeam said:

You can solve your equation online here: https://www.symbolab.com/solver/equ...t(A-Y\right)=\log_{10}\left(A\right)-C\cdot X

This is assuming you are using base 10 for your log function. If it is natural log, use this:

https://www.symbolab.com/solver/equ...-Y\right)=\log_{e^{1}}\left(A\right)-C\cdot X

Click to expand...

So that's what I got:

I'm pretty sure I'm doing something wrong, I've never done anything like that before.

 

Upvote 0

H

hyperchondriac

New Member

Joined

Jun 25, 2017

Messages

13

• Jun 25, 2017

• #6

It doesn't look like your data is anything like the model you are trying to fit to. There appears to be only a slight positive correlation between x and y, and no evidence of a logarithmic relationship.

Are you using the right model for the data?

HC

 

Upvote 0

P

phmalu

Board Regular

Joined

Jun 21, 2017

Messages

52

Office Version

1. 2019

Platform

1. Windows

• Jun 25, 2017

• #7

hyperchondriac said:

It doesn't look like your data is anything like the model you are trying to fit to. There appears to be only a slight positive correlation between x and y, and no evidence of a logarithmic relationship.

Are you using the right model for the data?

HC

Click to expand...

Yes, I'm sure it's the right model.
My data is from a field experiment, involving plants, soil and fertilization. I don't expect the correlation it to be perfect.
Those I've sent before are the "raw" values, I haven't transformed them to log or anything like that.

 

Upvote 0

M

mjbeam

Well-known Member

Joined

Apr 6, 2002

Messages

756

• Jun 25, 2017

• #8

Is the data you provided just a sample or is that all of it?

 

Upvote 0

H

hyperchondriac

New Member

Joined

Jun 25, 2017

Messages

13

• Jun 25, 2017

• #9

I'm not an expert in your field but it looks to me like you don't have enough data to fit to the model.

Either way - do you have a problem with the maths, or how to implement it in Excel?

HC

 

Upvote 0

P

phmalu

Board Regular

Joined

Jun 21, 2017

Messages

52

Office Version

1. 2019

Platform

1. Windows

• Jun 25, 2017

• #10

I have 16 points for each experiment. And yes, I'm aware I have too many points in the middle of the regression line and very few high/low values. This is completely normal when you're working with plants ans soil. I don't expect the R² to be excellent. At this point for me it's more important to fit that model to my data than fitting another model that represents better what I have.
I can deal with the maths, my biggest problem is how to plot it using excel.

 

Last edited: Jun 25, 2017

Upvote 0