In this case, you'd need to perform the followings in order to extrapolate backward:
(1) Import data, e.g.:
D = readtable('https://www.mathworks.com/matlabcentral/answers/uploaded_files/717269/PR1C_dailyvalues.xlsx')
(2) Select rows where nans are not present:
IDX = isnan(D.PR1C);
(3) Convert the dates to numbers:
F1 = D(IDX==1,1).Date; F1D = datenum(F1); F1D=F1D-F1D(1)+1;
F2 = D(IDX==0,1).Date; F2D = datenum(F2); F2D=F2D-F2D(1)+1;
Data = D(IDX==0,2).PR1C;
(4) Find a fit model:
FM = fit(F2D,Data,'poly3'); % Cubic polynomial
(5) Compute the extrapolation values:
FIT_Data = polyval(FM, F1D);
% etc.
