Help with moving percentile based on current information

[hhchenfx]

Hi Everyone,

I have a data with 2 variables value and N (N is the lag step used below)

For each record (i), I have to calculate the Change= value – lagN(value) for the prior 10 record. The N step is the N of current record.

So the below is how the calculation is done for record  11^th (with lag step =2). the 3^rd column is lag2 value, the 4^th column is the change.

Then I have to find the percentile of current change among these prior 10 prior change value.

(the most challenging part for me is to create a sub group for each record to do the rank, In practice, I have to find the percentile based on the prior 500 record)

Could you please help me to do that?

Thank you so much.

HHC

  2 2  .

  3 3  .

  9 4 2  9-2=7

  4 6 3 4-3=1

  5 8 9

  6 9 4

  7 1 5

  8 5 6

  6 2 7

  7 8 8

  8 2 6

--------------------------

data have;

  input value N;

  datalines;

  2 2

  3 3

  9 4

  4 6

  5 8

  6 9

  7 1

  8 5

  6 2

  7 8

  8 2

  3 4

  9 6

  0 9

  6 4

  4 8

  9 9

  8 3

  7 3

  4 4

  5 7

  4 5

  32 2

  21 1

  23 2

  4 3

  5 4

  6 6

  7 8

  8 2

;run;

[Ksharp]

Can't use PCTL Function to get it ? What does your output look like ?

[hhchenfx]

Hi Xia,

For each record, in the output we should have a new percentile collumn.

the difficulty for me is that the percentile of record ( i ) is built on the difference with lag step ( i )

I will create a spreadsheet when i get my laptop with me.

Thank you,

HHC

[ballardw]

You also may want to look at the DIFn function. Does the Lag and subtraction as one function.

And you only need the first variable from what you are showing.

If I understand what you may be looking for:

data want;

     set have;

     col2=dif2(n);

     col3=dif3(n);

     col4=dif4(n);

     col5=dif5(n);

     col6=dif6(n);

     col7=dif7(n);

     col8=dif8(n);

     col9=dif9(n);

     col10=dif10(n);

run;