matlab - 如何在 MATLAB 中一次性使用同一代码中的两个数据集？

Question

我有一个运行平稳的预先指定的模型。问题是我有两个相同模型的数据集。两个数据集仅根据行发生变化：例如，数据集 1 从 1 变为 60，数据集 2 从 61 变为 115。所有变量都相同。我想避免两次计算相同的代码。我宁愿把它写得很好，然后一口气计算出来。

我会给你一个带有我的模型的示例数据集：

data = rand(115,5)

Y_data = data(1:60, :) % dataset 1
Y_data = data(61:115, :) % dataset 2

% This is the model that runs nicely on dataset Y_data. I wanted to avoid to run the model twice,
% first with Y_data from row 1 to 60 and then from row to 61 to 100. I would like to do it in one shot
% the code for the model is fully automated so it's just a matter of making it work first on dataset 1 and then
% on dataset 2 in one unique code
T = size(Y_data,1);
P  = 3; % number of lags used in LP for controls
H_min = 1; 
H_max = 25; 
y  = Y_data(:,1); % endogenous variable
x  = Y_data(:,2); % shock 
w  = lagmatrix(Y_data(:,[3:5]), 1:P ); 
newData = cat(2, y, x, w)
% Remove missings from data
newData(any(isnan(newData), 2), :) = [];
% Re-declare variables after removing missings
y  = newData(:,1); % endogenous variable
x  = newData(:,2); % shock
w = newData(:,3:size(newData,2)); % control variables and lags
r = 3; 
lambda = 10000; 
slp    = locproj(y,x,w,H_min,H_max,'smooth',r,lambda); 
%% Cross-Validation Choice of Lambda
slp = locproj(y,x,w,H_min,H_max,'smooth',r,0.01);
lambda = [1:10:1000] * T;
slp    = locproj_cv(slp,5,lambda);
lambda_opt = lambda( min( slp.rss ) == slp.rss );
%% Confidence Intervals
r      = 3;
slp    = locproj(y,x,w,H_min,H_max,'smooth',r,lambda_opt); 
slp    = locproj_conf(slp,H_max,lambda_opt/2);

我认为可以解决这个问题的是使用 if/else，例如：

% This is wrong but it gives you an idea of what I was trying to do and get
% trying to tell MATLAB, fun the code first from dataset 1 (row 1:60) and then the same on dataset 2 (from row 61:115)

k = 1:60

if  k == 1

Y_data = Y_data;
    
else

       Y_data = data(61:115, :);
end

% model code as above just here - not to make it too long
% the output therefore should save both results for dataset1 and dataset2

我被卡住了，无法继续前进。谁能帮我？这会让我很开心。

非常感谢！

标签： matlabloopsfor-loopif-statement

Answer 1

有很多方法可以实现这一目标。这里有两个简单的。

做一个功能dataAnalysis(Y_data)，简单地做

[slp,lambdaOpt] = dataAnalysis(data(1:60,:))
[slp2,lambdaOpt2] = dataAnalysis(data(61:115,:))

你的功能可能看起来像

function [slp,lambdaOpt] = dataAnalysis(data)
% This is the model that runs nicely on dataset Y_data. I wanted to avoid to run the model twice,
% first with Y_data from row 1 to 60 and then from row to 61 to 100. I would like to do it in one shot
% the code for the model is fully automated so it's just a matter of making it work first on dataset 1 and then
% on dataset 2 in one unique code
T = size(Y_data,1);
P  = 3; % number of lags used in LP for controls
H_min = 1; 
H_max = 25; 
% etc. etc...
% Add in whatever output variables are important

或者，使用您尝试过的循环：

indices = {1:60, 61:115};
for k = 1:2
    Y_Data = data(indices{k},:);

    % your model code operating on Y_data
    % store any results here as either variable(k) (scalar data)
    % or variable {k} (non-scalar data)
end