# Delete rows from and interpolate

2 ビュー (過去 30 日間)
Damith 2015 年 2 月 25 日
コメント済み: Star Strider 2015 年 2 月 25 日
Hi,
Is there a way to implement in MATLAB to check first row and end row and then select 3 hr intervals rows from column 4 from table A (e.g. 0,3,6,9,12,15,18,21) and if not available include the 3 hr time steps and linear interpolate column 5 and 6. The new table should look like as shown in B. The "value" in col 5 and 6 should be obtained by linear interpolation. (See the test.xls attached)
A =
2000 1 24 4 -11.5 99.8
2000 1 24 6 -11.49 100.19
2000 1 24 10 -11.5 101
2000 1 24 12 -11.67 102.02
2000 1 24 16 -11.7 102.3
2000 1 24 18 -11.92 102.99
2000 1 24 22 -12.1 103.5
2000 1 25 0 -12.29 103.97
2000 1 25 4 -12.6 104.8
2000 1 25 6 -12.65 105.3
2000 1 25 10 -13 106.2
2000 1 25 12 -12.88 106.24
2000 1 25 16 -13.1 107.1
2000 1 25 18 -13.06 107.09
2000 1 25 22 -13.4 108.3
2000 1 26 0 -13.51 108.51
B =
2000 1 24 6 -11.49 100.19
2000 1 24 9 value value
2000 1 24 12 -11.67 102.02
2000 1 24 15 value value
2000 1 24 18 -11.92 102.99
2000 1 24 21 value value
2000 1 25 0 -12.29 103.97
2000 1 25 3 value value
2000 1 25 6 -12.65 105.3
2000 1 25 9 value value
2000 1 25 12 -12.88 106.24
2000 1 25 15 value value
2000 1 25 18 -13.06 107.09
2000 1 25 21 value value
2000 1 26 0 -13.51 108.51
Any help is highly appreciated.

サインインしてコメントする。

### 採用された回答

Star Strider 2015 年 2 月 25 日
This works:
DN = datenum([d(:,1:4) repmat([0 0], size(d,1), 1)]); % Convert All To ‘datenum’
D0 = floor(DN(1)); % Start Date & Time
DE = ceil(DN(end)); % End Date & Time
DT = 1/24; % Days/Hour
DV = D0 + cumsum([0; ones(round(24*(DE-D0)),1)*DT]); % All Hours
DV = DV(1:3:end-1); % q3 Hours
DC = unique([DN; DV]); % Combine Date Vectors
DataIntrp = interp1(DN, d(:,5:6), DC); % Interpolate
DateMtx = datevec(DC); % Date Vectors
DataResult = [DateMtx(:,1:4), DataIntrp];
DataResult = DataResult(~isnan(DataResult(:,5)),:); % Output Eliminating Extrapolations
CheckResult = DataResult([1:10 end-9:end],:) % Check Result (Temporary)
The ‘CheckResult’ line is there to look at the first and last 10 entries. It doesn’t match your ‘B’ matrix, but it does match your description (and your question in my not yet Accepted Answer to your previous Question). It could likely be made more efficient, but if you only need to run it once for each data set, this is likely sufficient.
##### 2 件のコメント表示非表示 1 件の古いコメント
Star Strider 2015 年 2 月 25 日
My pleasure!

サインインしてコメントする。

### その他の回答 (1 件)

Andrei Bobrov 2015 年 2 月 25 日
n = find(rem(x(:,4),3)==0);
x1 = x(n(1):n(end),:);
t0 = num2cell(x1(:,1:4),1);
t = datenum(t0{:},0,0);
t2 = (t(1):3/24:t(end))';
[y,m,d,h] = datevec(t2);
out = [y,m,d,h,interp1(t,x1(:,end-1:end),t2)];
##### 1 件のコメント表示非表示 なし
Damith 2015 年 2 月 25 日
Thanks Andrei. Your code produce the exact output I want. I appreciate it.

サインインしてコメントする。

### Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by