Re: [Gretl-users] Correcting for inflation

On Sat, 11 Apr 2009, Raz Lev wrote: 400,000 monthly CPI figures would be 33,333 years' worth, so I assume you can't have that. Even 400,000 daily observations would be about 1000 years' worth, so I guess you don't have that either. My best guess, then, is that you have (a) a panel with a large number of "individuals" observed in each of several months or years, plus (b) a plain time series of CPI figures, covering all the time periods in the panel sample, and you want to expand the CPI data into panel form, so that (e.g.) the CPI value for 1980:01 gets written into each slot in the big dataset that records an individual's characteristics in 1980:01. If my interpretation is right, this is a somewhat fiddly task. Here's one way to do it (I suspect there may be a more elegant way). Let's say I have these CPI data in cpi.txt: date cpi 1980:01 100 1980:02 101 1980:03 103 1980:04 105 1980:05 106 and let's say I want to create a data file containing 30 stacked copies of this series for 30 panel units. Then <script> open cpi.txt # contains time-series data scalar n = 30 # assuming 30 panel units scalar T = $nobs # number of CPI observations scalar nT = n * T # target number of observations matrix m = zeros(nT, 1) scalar k = 1 loop n --quiet loop j=1..T --quiet m[k] = cpi[j] k++ endloop endloop # create full-length dataset, preserving the matrix nulldata nT --preserve # convert matrix to series series bigcpi = m # define the new dataset as a panel setobs T 1:1 --stacked-time-series # check that our expanded series looks OK print bigcpi --byobs # save as gretl data file store bigcpi.gdt bigcpi </script> The new data file bigcpi.gdt can now be appended (File/Append data/Standard format) to an existing panel data file in stacked time series format. Allin Cottrell