# How to Calculate linear trend for each column of a matrix

97 views (last 30 days)
Shakir Hussain on 23 Nov 2018
Answered: Star Strider on 23 Nov 2018
Can we calculate the linear trend of each column of a matrix without drawing any line,
For example I want to calculate the linear trend for the column a,b,c,d,e of the attached data
Thank you for any tips or help

Star Strider on 23 Nov 2018
Try this:
B = [(1:size(D,1))', ones(size(D,1),1)] \ D
producing:
B =
-0.7307 -0.6702 -0.5743 -0.5264 -0.3016
38.1515 33.7212 35.0423 35.3009 35.6209
The first row are the slopes and the second row are the intercepts. The columns are for ‘a’ through ‘e’ respectively.

Luna on 23 Nov 2018
Hi,
I think you can use polyfit simply for this problem.
Here is an example code. You can apply it for the other columns aswell.
columnNames = txt(1,:);
columnNames(1) = [];
for i = 1:numel(columnNames)
colData.(columnNames{i}) = num(:,i);
end
x = 1:12;
linearCoefficients = polyfit(x,colData.a',1); % First degree linear fit
yFit = polyval(linearCoefficients, x);
% Apply it for colData.b and c .. so on.
%% To see data and line
figure;
plot(x,colData.a','bo');
hold on;
plot(x,yFit,'-r');

Luna on 23 Nov 2018
Btw, I assumed your x data from 1 to 12. You can choose any other column as reference.
Shakir Hussain on 23 Nov 2018
dear luna, thanks for your response
yes you rightly assumed it. Could we deal it a matlab matrix without having column names? and have R and P value of all Y columns.
Luna on 23 Nov 2018
Ah, that columnNames I just only used to extract the xls file easily.
You can always use below xlsread function outputs as you wish.
num gives you the only numerical matrix, txt only gives you the non-numeric text values, raw is the whole data as a cell array.
just get num and take each column of it one by one.