Importing timeseries datasets to MATLAB (all values are displayed as NaN) - file-io

Question

I am stuck trying to run an economic model using MATLAB - at the data importing part. For most of my code I'm using a freeware toolbox called IRIS.
I have quarterly dataset with 14 variables and 160 datapoints. Essentially the dataset is a 15X161 matrix- including the dates(col1) and variable names(B1:O1).
The command used for uploading data on IRIS is
d = dbload('filename.csv')
but this isn't working. Although MATLAB is creating a 1X1 array called d and creating fields under it (one for each variable). All cells display NaN - not a number.
Why is this happening?
I checked the tutorials on the IRIS toolbox website and tried running and loading a sample dataset from there using this command, but it leads to the same problem. Everywhere I checked- including MATLAB help, this seems to be the correct command to use when using IRIS, but somehow it isn't working.
I also tried uploading the data directly using MATLAB functions and not IRIS. The command I'm using is:
d = dataset('XLSFile','filename.xls','ReadVarNames', true).
Although this is working, and I can see all the variable names, but MATLAB can't read the dates. I tried xlsread and importdata as well, but they don't read the variable names. Is there any way for me to upload the entire Excel sheet with the variable names and dates?
It would be best if I could get the IRIS command to work, since the rest of my code would be compatible with that.
The dataset looks somewhat like this..
HO_GDP HO_CPI HO_CPI HO_RS HO_ER HO_POIL....
4/1/1970 82.33 85.01 55.00 99.87 08.77
7/1/1970 54.22 8.98 25.22 95.11 91.77
10/1/1970 85.41 85.00 85.22 95.34 55.00
1/1/1971 85.99 899 8.89 85.1

Answer 1

You can use the TEXTSCAN function to read the CSV file in MATLAB:
%# some options
numCols = 15; %# number of columns
opts = {'Delimiter',',', 'MultipleDelimsAsOne',true, 'CollectOutput',true};
%# open file for reading
fid = fopen('filename.csv','rt');
%# read header line
headers = textscan(fid, repmat('%s',1,numCols), 1, opts{:});
%# read rest of data rows
%# 1st column as string, the other 14 as floating point
data = textscan(fid, ['%s' repmat('%f',1,numCols-1)], opts{:});
%# close file
fclose(fid);
%# collect data
headers = headers{1};
data = [datenum(data{1},'mm/dd/yyyy') data{2}];
The result for the above sample you posted (assuming values are comma-separated):
>> headers
headers =
'HO_GDP' 'HO_CPI' 'HO_CPI' 'HO_RS' 'HO_ER' 'HO_POIL'
>> data
data =
7.1962e+05 82.33 85.01 55 99.87 8.77
7.1971e+05 54.22 8.98 25.22 95.11 91.77
7.198e+05 85.41 85 85.22 95.34 55
7.1989e+05 85.99 899 8.89 85.1 0
Note how in the last line of the code we convert the date column to serial date number, so that we can store the entire data in one numeric matrix. You can always go back to string representation of dates using DATESTR function:
>> datestr(data(:,1))
ans =
01-Apr-1970
01-Jul-1970
01-Oct-1970
01-Jan-1971