How can I easily extract element from a big matrix?

I have a bigg matrix [365x24] whic represtens a specific property for each hour and day of the year.

It can be read like this:

The first line has of 24 elements whit elements whic are corresponding to the 00:00 to 23:00 of Monday; the second is Tusday and so on with the Seventh that is Sunday and then it restart with the following weekt.

The first 31 row represent the days of January the following 28 rows rapresents the days of february and so on...

I would like to extrat some specific intervals of data for each month and for the whole year, that are

F1--the elements fom 8:00- to 19:00 (8 am to 7 pm) for each moday to friday interval

F2--the elements and 20:00 to 23:00 for each Monday to Friday interval and Saturday all day from 0:00 to 23

F3-- The elements from 00:00 to 7:00 (midnight -7 am) for each Monday to Saturday interval; and al Sundays all day fro 00:00 to 23

The intervals are summarized in figure

I tried to do it mannualy but it's a messy and slow method, how can i do it quickly?

%For example for F1
A=randi(365,24);
F1m=[A(1:5,8:19);A(8:12,8:19);A(15:19,8:19); A(22:26,8:19);A(29:33,8:19); A(36:40,8:19);
    A(43:47,8:19);A(50:54,8:19);A(57:61,8:19);A(CYear(64:68,8:19);A(71:75,8:19); A(78:82,8:19)...   ];
%and the sum
F1=sum(sum(F1m));

[KSSV]

How do you have the data? Is it in excel? Attach your data. This can be achioeved by converting your date, times into datetime class and then by searching what you want.

[Pietro Fiondella]

It's not excel is an actual matrix in matlab founded by computation on matlab, and it's only coefficients, only the position has meaning. meaby it can be done the inverse, i mean taking that on excel

[KSSV]

Attach your data.

[Pietro Fiondella]

well I try to then.

ProfileRh=[0.246 0.280 0.286 0.301 0.302 0.330 0.397 0.478 0.607 0.651 0.595 0.495 0.546 0.594 0.609 0.608 0.628 0.866 0.944 0.988 0.882 0.831 0.745 0.616];
DayC_R=[ 0.978; 0.992; 0.997; 0.997; 0.995; 0.963; 0.881];
MonthC_R=[0.999 0.992 0.939 0.850 0.810 0.803 0.796 0.792 0.821 0.929 0.990 1];

% Then the calculation are
MatCY_Rm=[MCoeff(ProfileRh,DayC_R,MonthC_R,1,31);MCoeff(ProfileRh,DayC_R,MonthC_R,2,28);MCoeff(ProfileRh,DayC_R,MonthC_R,3,31);MCoeff(ProfileRh,DayC_R,MonthC_R,4,30);MCoeff(ProfileRh,DayC_R,MonthC_R,5,31);MCoeff(ProfileRh,DayC_R,MonthC_R,6,30);MCoeff(ProfileRh,DayC_R,MonthC_R,7,31);MCoeff(ProfileRh,DayC_R,MonthC_R,8,31);MCoeff(ProfileRh,DayC_R,MonthC_R,9,30);MCoeff(ProfileRh,DayC_R,MonthC_R,10,31);MCoeff(ProfileRh,DayC_R,MonthC_R,11,30);MCoeff(ProfileRh,DayC_R,MonthC_R,12,31)];
 MatCY_R=MatCY_Rm/sum(sum(MatCY_Rm));
%  Then MatCY_R is the matrix from where i should extract the data 
 MatCY_R

MatCY_R = 365×24

1.0e-03 * 0.0482 0.0549 0.0561 0.0590 0.0592 0.0647 0.0778 0.0937 0.1190 0.1276 0.1166 0.0970 0.1070 0.1164 0.1194 0.1192 0.1231 0.1697 0.1850 0.1936 0.1729 0.1629 0.1460 0.1207 0.0489 0.0557 0.0568 0.0598 0.0600 0.0656 0.0789 0.0950 0.1206 0.1294 0.1183 0.0984 0.1085 0.1181 0.1210 0.1208 0.1248 0.1721 0.1876 0.1964 0.1753 0.1652 0.1481 0.1224 0.0491 0.0559 0.0571 0.0601 0.0603 0.0659 0.0793 0.0955 0.1212 0.1300 0.1188 0.0989 0.1090 0.1186 0.1216 0.1214 0.1254 0.1730 0.1885 0.1973 0.1762 0.1660 0.1488 0.1230 0.0491 0.0559 0.0571 0.0601 0.0603 0.0659 0.0793 0.0954 0.1212 0.1300 0.1188 0.0988 0.1090 0.1186 0.1216 0.1214 0.1254 0.1729 0.1885 0.1973 0.1761 0.1659 0.1488 0.1230 0.0490 0.0558 0.0570 0.0600 0.0602 0.0657 0.0791 0.0952 0.1209 0.1297 0.1185 0.0986 0.1088 0.1183 0.1213 0.1211 0.1251 0.1725 0.1881 0.1968 0.1757 0.1656 0.1484 0.1227 0.0474 0.0540 0.0551 0.0580 0.0582 0.0636 0.0765 0.0921 0.1170 0.1255 0.1147 0.0954 0.1053 0.1145 0.1174 0.1172 0.1211 0.1669 0.1820 0.1905 0.1700 0.1602 0.1436 0.1188 0.0434 0.0494 0.0504 0.0531 0.0532 0.0582 0.0700 0.0843 0.1070 0.1148 0.1049 0.0873 0.0963 0.1047 0.1074 0.1072 0.1107 0.1527 0.1664 0.1742 0.1555 0.1465 0.1314 0.1086 0.0481 0.0548 0.0560 0.0589 0.0591 0.0646 0.0777 0.0935 0.1188 0.1274 0.1164 0.0969 0.1068 0.1162 0.1192 0.1190 0.1229 0.1695 0.1847 0.1933 0.1726 0.1626 0.1458 0.1205 0.0488 0.0556 0.0568 0.0597 0.0599 0.0655 0.0788 0.0949 0.1205 0.1292 0.1181 0.0982 0.1084 0.1179 0.1209 0.1207 0.1246 0.1719 0.1873 0.1961 0.1750 0.1649 0.1478 0.1222 0.0491 0.0558 0.0570 0.0600 0.0602 0.0658 0.0792 0.0953 0.1210 0.1298 0.1186 0.0987 0.1089 0.1184 0.1214 0.1212 0.1252 0.1727 0.1882 0.1970 0.1759 0.1657 0.1486 0.1228

 
 %%% To obtain the matrix a function MCoeff is required me and it wors like this
function MatCoeff = MCoeff(p,d,Mc,m,n)
%WeekCoeff gives the coefficient of week by interpolation of data
% p=Profile, d=DayCm Mc=MonthCoeff, m=month number n=number of interval (n=4 weeks each month, or n=# of days in a month)
A=p.*d;
if n>28
Ac=[A;A;A;A;A(1:n-28,:)];
else  
    Ac=[A;A;A;A];
end
M= linspace(n,n+1,2);
y=linspace(M(1),M(end),n*(numel(M)-1)+1) ;
if m<12
z=Mc(m:m+1);
else
 z=[Mc(1,m),Mc(1,1)];
end
 W=interp1(M,z,y);
WCoeff=W((1:length(W)-1));
MatCoeff1=(Ac.*WCoeff');
MatCoeff=(Ac.*WCoeff')/(sum(sum(MatCoeff1)));
end

[Pietro Fiondella]

As i said I obtain here this big matrix in whic elements have meaning in respect of their position (Day, Hour, week, month) and I shouldbe able to "select" and operate with some specific group of them.

%For example for the range F1 that goes from Monday to Friday 8-19 (column 9 to 20 counting from 0:00) 
% I'm trying to select multiple parts

    F1=sum(sum(Matrix(:,9:20));
%   Here I should not take all the rows,
% but instead  avoid 2 rows (Saturaday, Sunday) each 5(Monday-Friday)

[dpb]

Don't make it so hard...use the paper clip icon and attach the .mat file.

It also probably would be simpler to turn the time variables into datetimes (assign an arbitrary year) and convert to a timetable so can use the features of retime (although I've not tried to manually recreate a similar data structure to test; that's why you should attach yours...folks are far more likely to poke at a problem if there's something there to play with rather than have to also make it up from scratch...)

[Pietro Fiondella]

I don know if i get it, but extract the matrix as .xls and attached here should it help?

[Adam Danz]

I assume you have a 1x24 vector that defines the time for each column or at least you the time of the first column and the duration of each column which could be used to compute the 1x24 time vector.

I also assumes you have a 365x1 vector of dates that define each row or a starting date and the duration of each row that could be used to compute the date vector.

Once you have those values, you could create a timetable as @dpb suggested above. Once you have a timetable, your task becomes much simpler.

[Pietro Fiondella]

No i havent but i could create them.,But i dont know what i shlod do with them then.

Can someone make an example with an easier matrix idk [5x5]? I will accept it as an anwser.

Answer 1

[dpb]

Yeah, I'd convert to timetable and reorganize -- something like this

Time=datetime(2022,1,1,[0:365*24-1].',0,0);         % create the time vector hourly; arbitrary nonleapyear
M=readmatrix('Matrix.xlsx').';                      % read data, arrange by hour
ttM=timetable(Time,M(:),'VariableNames',{'V'});     % create base timetable
ttM.Time.Format='HH';                               % just show the hours
Class=categorical("F"+[1:3].');                     % your class grouping variable      
ttM=addvars(ttM,repmat(Class(3),height(ttM),1),'Before','V','NewVariableNames','Class');    % Class 3 is largest
isF2=isbetween(timeofday(ttM.Time),hours(7),hours(23))&(~matches(day(ttM.Time,'name'),'Sunday'));   % next subset
ttM.Class(isF2)=Class(2);
isF1=isbetween(timeofday(ttM.Time),hours(8),hours(19))&~isweekend(ttM.Time);    % smallest
ttM.Class(isF1)=Class(1);

The above gives

>> head(ttM,10)
ans =
  10×2 timetable
    Time    Class       V   
    ____    _____    _______
    00       F3      0.70825
    01       F3      0.74999
    02       F3      0.79296
    03       F3      0.85226
    04       F3       0.9795
    05       F3      0.99869
    06       F3        1.207
    07       F2       1.4139
    08       F2       1.6649
    09       F2       1.6355
>> 

Now use the Class variable as the grouping variable and the rest is piece o' cake w/ retime() or rowfun() or grpstats() or ...

I'll leave as "exercise for Student" adding DOW names and weekend or whatever other pretty variables care to add...

[Pietro Fiondella]

Thank you dpb. I am not totally sure of what have you did but I'll study this

[dpb]

Turned your data on its head as linear vector versus time and created the time vector to go with it explicitly instead of trying to infer something by position.

Study the sections in the doc on tables and timetables to begin to get an idea of their flexibiliity and power -- don't overlook the 'Functions' link at the top of the specific doc pages -- it has a comprehensive list of all the functions tailored specifically for manipulating the data class...and, of course, the section on addressing data in a table is indispensible.

[dpb]

>> groupsummary(ttM,'Class','sum')
ans =
  3×3 table
    Class    GroupCount    sum_V 
    _____    __________    ______
     F1         3120         4928
     F2         2201       3033.5
     F3         3439       3434.5
>>