WOuld Analytic Windowing Clause help solve this?

Question

I've been using some basic analytic functions for a little while now. Trying to do something that I don't know if it's syntactically legal...or if maybe I'm just approaching it the wrong way.
I have a situation like this...
with t1 as
(SELECT 1 AS id , 1 AS value_1, 'Y' valid_1 , 12 AS value_2 , 'N' valid_2 , 38 AS value_3 , 'N' valid_3 , 45 AS value_4 , 'Y' valid_4 FROM dual
UNION ALL
SELECT 2 AS id , 3 AS value_1, 'N' valid_1 , 7 AS value_2 , 'Y' valid_2 , 13 AS value_3 , 'Y' valid_3 , 19 AS value_4 , 'N' valid_4 FROM dual)
select
  id,
  pivot,
  pivoted_value,
  pivoted_valid,
  max(pivoted_value) over (partition by id
                           order by pivot
                          ) as analytic_expr
from (
select
  t1.id,
  t2.pivot,
  case t2.pivot
    when 1 then t1.value_1
    when 2 then t1.value_2
    when 3 then t1.value_3
    when 4 then t1.value_4
  end as pivoted_value,
  case t2.pivot
    when 1 then t1.valid_1
    when 2 then t1.valid_2
    when 3 then t1.valid_3
    when 4 then t1.valid_4
  end as pivoted_valid
from t1, (select rownum pivot from dual connect by rownum < 5) t2
        ID      PIVOT PIVOTED_VALUE P ANALYTIC_EXPR
         1          1             1 Y             1
         1          2            12 N            12
         1          3            38 N            38
         1          4            45 Y            45
         2          1             3 N             3
         2          2             7 Y             7
         2          3            13 Y            13
         2          4            19 N            19...which get's me close to what I'm looking to do. But rather than having a running max, I'd need it to increment only when the value is "valid". So for row 2 above, the value of analytic_expr should stay 1 for rows 2,3. Row 5 should be null (or some other defautable value), and the last row should stay at 13.
Does <where pivot_value = 'Y'> translate into some useable windowing clause, or so I have to do something a little less elegant?

Answer 1

Mentioning that values 1-4 are always increasing would probably made my statement sensible.
If pivoted_valid(X) = 'Y' (column name name got copped to "P" for some reason)
Then pivoted_value(x) ....which will also be the current max if they're ordered in
Else the max(pivoted_value(<x))
This is how I would want the results to look...
ID   PIVOT    VAL   VALID    CURRENT_MAX_VAL
1     1     1     Y     1
1     2     12     N     1    --since not valid, no change to prior
1     3     38     N     1     --since not valid, no change to prior
1     4     45     Y     45
2     1     3     N     nulll    --since not valid and no prior row
2     2     7     Y     7
2     3     13     Y     13
2     4     19     N     13   --since not valid, no change to priorI know I could probably do this with a CASE/LAG combo of some sort. I was just wondering how fancy a person could get with windowing.