How to import and create time series data frames in an efficient way?

Ha! -- A bug! "Corrected" version inline below:
Bert Gunter

On Thu, Nov 14, 2019 at 8:10 PM Bert Gunter <bgunter.4567 at gmail.com>
wrote:

Brute force approach, possibly inefficient:

1. You have a vector of file names. Sort them in the appropriate (time)
order. These names are also the component names of all the data frames in
your list that you read in, call it yourlist.

2. Create a vector of all the unique ticker names, perhaps by creating a
vector of all the names and then unique() -ing it. Call this vector snames
with n.names in it. It will probably have length several hundred at least I
assume.

3. Suppose the  6 columns of data of each data frame that you want are
named cnames = c("stocknames","Open", "High", "Low", "Close", "Volume").

4. You could proceed as you suggested, but it would likely be more
efficient, since all data that you want are numeric, to create a 3D array
of NA's via:

yourdat <- array(dim = c(n.dates, n.names, 5), dimnames = list(NULL,
snames, cnames[-1]))

5. Then just loop  through your list of files and use indexing to fill in
the columns x category slices for each date. Stocks that are missing will
be NA automatically. e.g. (warning: UNTESTED):

For date "d", let df be the data frame from date "d" in your list, i.e.

df <- yourlist[["d"]][, cnames]
## Note The order of the listed stocks in the "stocknames" column can be
different from frame to frame of your master list.

Then fill in the flat for the dth date (i.e. dth row) in your array by:

## corrected line here:

yourdat[ ,df[ ,"stocknames"], cnames[-1] <- as.matrix(df[ ,-1]) ## need
to omit the column names so it converts to numeric matrix

## need to get the names of the stocks in the "stocknames" column in the
order they appear in df.

This should fill in  the values of the 2nd and 3rd dimensions of the
array for all the stocks on the dth date with the data for each stock in
the data frame matched to the appropriate column in the array.

The entire loop will give all dates for all stocks and all categories
with NA's for missing days. (*IF IT WORKS!*)
You may need to modify this sightly if, for example, your stock names are
row names rather than a field in your data frame. I leave such adjustments
to you.

Note again that this is fairly elementary with just arrays and indexing.
Basic tutorials should tell you about all of this. Also, when plotting,
you'll have to convert your dates to suitable date-time format.

Cheers,
Bert




On Thu, Nov 14, 2019 at 4:55 PM Nhan La <lathanhnhan at gmail.com> wrote:

Hi Bert,

I've attempted to find the answer and actually been able to import the
individual data sets into a list of data frames.

But I'm not sure how to go ahead with the next step. I'm not necessarily
asking for a final answer. Perhaps if you (I mean others as well) would
like a constructive coaching, you would suggest a few key words to look at?

Sorry for the HTML thing, this is my first post. I'll do better next
times.

Thanks,
Nathan



On Fri, Nov 15, 2019 at 11:34 AM Bert Gunter <bgunter.4567 at gmail.com>
wrote:

So you've made no attempt at all to do this for yourself?!

That suggests to me that you need to spend time with some R tutorials.

Also, please post in plain text on this plain text list. HTML can get
mangled, as it may have here.

-- Bert
"The trouble with having an open mind is that people keep coming along
and sticking things into it."
-- Opus (aka Berkeley Breathed in his "Bloom County" comic strip )


On Thu, Nov 14, 2019 at 4:11 PM Nhan La <lathanhnhan at gmail.com> wrote:

I have many separate data files in csv format for a lot of daily stock
prices. Over a few years there are hundreds of those data files, whose
names are the dates of data record.

In each file there are variables of ticker (or stock trading code),
date,
open price, high price, low price, close price, and trading volume. For
example, inside a data file named 20150128.txt it looks like this:

FB,20150128,1.075,1.075,0.97,0.97,725221
AAPL,20150128,2.24,2.24,2.2,2.24,63682
AMZN,20150128,0.4,0.415,0.4,0.415,194900
NFLX,20150128,50.19,50.21,50.19,50.19,761845
GOOGL,20150128,1.62,1.645,1.59,1.63,684835 ...................and many
more..................

In case it's relevant, the number of stocks in these files are not
necessarily the same (so there will be missing data). I need to import
and
create 5 separate time series data frames from those files, one each
for
Open, High, Low, Close and Volume. In each data frame, rows are
indexed by
date, and columns by ticker. For example, the data frame Open may look
like
this:

DATE,FB,AAPL,AMZN,NFLX,GOOGL,... 20150128,1.5,2.2,0.4,5.1,1.6,...
20150129,NA,2.3,0.5,5.2,1.7,... ...

What will be an efficient way to do that? I've used the following
codes to
read the files into a list of data frames but don't know what to do
next
from here.

files = list.files(pattern="*.txt") mydata = lapply(files,
read.csv,head=FALSE)

Thanks,

Nathan

Disclaimer: In case it's relevant, this question is also posted on
stackoverflow.

        [[alternative HTML version deleted]]