How to use LLSimpute in MatLab

I was trying to use the LLSimpute software ( http://www.cc.gatech.edu/~hpark/softwareMVE.html) in MATLAB. But got difficulties with lots of MATLAB's complain. It worked fine with the example data given with this software. I removed (in my data) the column and row names and changed missing values as requested in the program. Copied my data in the same directory where these LLSimpute programs are. Then, even tried to convert the csv file into a .mat file and trying to execute the first program,'main missing value code'(mpute_llsq_l2.m) like

E=impute_llsq_l2(0,1);

• Even after renaming my new matrix with missing values with the same file name as in the training data, miss0.mat.
• After loading the data ('load miss0.mat') i changed 'data' in the workspace of MATLAB into 'miss_matrix' - as this is needed by the program,mpute_llsq_l2.m.

But i am still getting an error message like the following

>> E=impute_llsq_l2(0,1);
??? Undefined function or method 'eq' for input arguments of type 'struct'.

Error in ==> impute_llsq_l2 at 69
if (f_rowaverage==1) | (m-length(find(miss_matrix==big))) < 400

I am new to MATLAB, can anyone please help me how I can use this LLSimpute software, impute_llsq_l2.m, to impute my missing data which is in csv file format?

Thank you

Answer 1

[Walter Roberson]

Exactly how did you load the data? The complete command?

If you had

miss_matrix = load('miss0.mat');

then miss_matrix would be returned as a structure that had fields whose names were the names of the variables in the .mat file.

You might instead need to

miss_load = load('miss0.mat');
miss_matrix = miss_load.miss_matrix;

[Daniel]

Hello Walter,

Thanks for your solution. But here is how I did try to load it and your solution when I apply it in MATLAB.

My way:

load('my_data.csv'); % this loads my data in to the MATLAB % workspace as 'my_data<743*743 double>'

miss_matrix = my_data;

Your way:

miss_load = load('miss0.mat'); % <1*1 struct>

miss_matrix = miss_load.miss_matrix;

??? Reference to non-existent field 'miss_matrix'.

But anyways, I just continued with just the 'miss_matrix' i got and MATLAB was still complaining about other error sources (which were not there when i used the sample data)and i tried to fix them by reading and looking for help from the internet. But finally i got a zero output. i.e an empty matrix!

I suspect that the other parameters like the 'minexp=2;'in line 58 0f 'impute_llsq_l2.m' need to be changed,like '%minexp=n*0.3;' as commented and correspondingly in 'kestimate_12.m'code for the plot in 'plot(xk, nrmse,'k^-'); axis([0 k_max 0 2.0]);'. But even then, I still not getting an output. Help, help, help....pls

Thank you

[Daniel]

Walter, what was your 'miss_matrix'? Did you assume mine or anything else?

Answer 2

[yaya wu]

Hello,do you solve this problem?

Answer 3

[yaya wu]

May I know that why can't I open this website" http://www.cc.gatech.edu/~hpark/softwareMVE.html"? Thanks for your reply!