2008R2 - cumulative based on count and additional subquery where CASE SUM

Question

Hi;
I have two queries where I'm struggling to get a cumulative and sum due to use of counts in the initial query;
First up cumulative where I suspect a cursor will be required.
SELECT TOP (100) PERCENT DATEPART(week, time_stamp) AS Week, COUNT(DISTINCT [user]) AS UnqCnt
FROM dbo.vw_ctrx_archive_ica_roundtrip_perf
WHERE (time_stamp BETWEEN GETDATE() - 120 AND GETDATE()) AND (dept_path LIKE 'All\XenApp Farms\AppName\%')
GROUP BY DATEPART(week, time_stamp)
ORDER BY week
This produces a unique count of users grouped by week number, However What I want to show is a cumulative total as each week goes by;
Week 1 100 unique users
By end of Week 2
220 unique users
By end of week 3 
250 unique users and so forth.
In this instance this level of reporting is only required for the next 10 - 12 weeks so I'm happy to go inefficient ala in the vain of;
SELECT TOP (100) PERCENT Week, UnqCnt,
(SELECT SUM(UnqCnt) AS Expr1
FROM dbo.vw_someview AS t2
WHERE (Week <= t1.Week)) AS RunningT
FROM dbo.vw_someview AS t1
ORDER BY Week
(The above isn't what I need as it simply totals each week as opposed to showing unique to date, but less complexity works for me!)
EDIT - Examples of source data and desired output. I've trimmed down to show just the required data.
Current input;
Current output;
What I actually want is a cumulative unique count by week to date, not per week.
So the next query is based on a count of unique users plus a CASE SUM showing split between two regions (UK and US). However there is a mismatch with the unique user count total due to potential multiple records existing for a user which is included in the
CASE SUMs. Again I know a subquery is necessary but I'm struggling to work it in conjunction with the CASE SUMS.
SELECT        TOP (100) PERCENT time_stamp, CONVERT(varchar(2), DATEADD(day, 0, time_stamp), 105) + '-' + CONVERT(varchar(2), DATEADD(hh, 0, time_stamp), 108) 
                         + ':00' AS Time, COUNT(DISTINCT [user]) AS Combined, 
                         SUM(CASE WHEN dbo.vw_ctrx_archive_session_perf.dept_path LIKE 'All\XenApp Farms\someservers\UK%' THEN 1 ELSE 0 END) AS UK, 
                         SUM(CASE WHEN dbo.vw_ctrx_archive_session_perf.dept_path LIKE 'All\XenApp Farms\someservers\US%' THEN 1 ELSE 0 END) AS US
FROM            dbo.vw_ctrx_archive_session_perf
WHERE        (dept_path LIKE 'All\XenApp Farms\someservers\%')
GROUP BY time_stamp, CONVERT(varchar(2), DATEADD(hh, 0, time_stamp), 108)
ORDER BY time_stamp
EDIT - Examples of source data and desired output. I've trimmed down to show just the required data.
Current input;
Current output;
As you can see the mismatch between Combined and region due to no logic with the CASE SUMS to only perform that SUM for a unique user.
Any help or pointers for either question would be very much appreciated.
Thanks, Paul.

Answer 1

The first query the cte returns one row for each customer who has a order into the @MinWeekCounts table variable.  With that customer id, the cte returns the first week that customer has an order.  So if customer number 1 has two orders in week
3, 5 orders in week 7 and one order in week 9, the cte will return one row for customer number 1 with the week set to 3.
Then the result of the cte (one row for each customer with the week that customer first had an order) is grouped by week and we get a count of the customers for that week.  So now we have for each week the count of the customers whose first order was
in that week.  So for week 1 we have the count of the customers who ordered in week 1.  For week 2, we have the count of the customers who ordered in week 2, but did not order in week 1.  For week 3, we have the count of the customers who ordered
in week 3, but did not order in either week 1 or week 2.  The query then inserts that result into the @MinWeeksCounts table variable.
So now we have a count for each week of the number of customers whose first order was in that week.  So the result we want is just the running total of those counts.  For example, for week three, you want the number of customers who ordered ordered
in week 3 or earlier.  That's just the number of customers who first ordered in week 1 + the number of customers who first ordered in week 2 plus the number of customers who first ordered in week 3.  So the running total for each week will give
you the result you want.  So we write the final query to give you the result by doing a running total of each week's count of the customers who first ordered in each week.
Tom