Hi all,
I wanted to check if anyone had any experience using predictors for 1) «carry forward» of values from one period to the following, or 2) copying cohort data forward?
As far as I can tell from some limited testing, 1) should work as it is now by setting sequential sample count to 1 and annual sample count to 0. You could then for example copy ending balance for May 2017 to starting balance June 2017 etc. (note: I don’t mean to start a discussion how to collect data on stock).
Moving forward cohorts would similarly work today, but only for 12 month cohorts as far as I can figure out. From what I understand, even 12-month cohorts are sometimes in reality reported after e.g. 14 months, and there are of course others (6 months, 24 months, 36 months). At least on the surface, it would seem adding support for specifying the sample period as X months back would be a relatively straightforward thing, and it would be very helpful for this purpose. I know people are using custom script for this type of thing today.
I am not sure about all cohort types, would have to look a bit more. I believe for 6 months cohort it might work like this (remember I haven’t tested).
Period type = Monthly
Annual sample count = 0
Sequential sample count = 6
Sequential skip = 5
Then it would exclude the first 5 months of data in relation to the sample selected while also going 6 months back.
I guess you could potentially do the same for the other cohorts as long as your skip identified how far back you went?
I wanted to check if anyone had any experience using predictors for 1) «carry forward» of values from one period to the following, or 2) copying cohort data forward?
As far as I can tell from some limited testing, 1) should work as it is now by setting sequential sample count to 1 and annual sample count to 0. You could then for example copy ending balance for May 2017 to starting balance June 2017 etc. (note: I don’t mean to start a discussion how to collect data on stock).
Moving forward cohorts would similarly work today, but only for 12 month cohorts as far as I can figure out. From what I understand, even 12-month cohorts are sometimes in reality reported after e.g. 14 months, and there are of course others (6 months, 24 months, 36 months). At least on the surface, it would seem adding support for specifying the sample period as X months back would be a relatively straightforward thing, and it would be very helpful for this purpose. I know people are using custom script for this type of thing today.
I am not sure about all cohort types, would have to look a bit more. I believe for 6 months cohort it might work like this (remember I haven’t tested).
Period type = Monthly
Annual sample count = 0
Sequential sample count = 6
Sequential skip = 5
Then it would exclude the first 5 months of data in relation to the sample selected while also going 6 months back.
I guess you could potentially do the same for the other cohorts as long as your skip identified how far back you went?
I wanted to check if anyone had any experience using predictors for 1) «carry forward» of values from one period to the following, or 2) copying cohort data forward?
As far as I can tell from some limited testing, 1) should work as it is now by setting sequential sample count to 1 and annual sample count to 0. You could then for example copy ending balance for May 2017 to starting balance June 2017 etc. (note: I don’t mean to start a discussion how to collect data on stock).
Moving forward cohorts would similarly work today, but only for 12 month cohorts as far as I can figure out. From what I understand, even 12-month cohorts are sometimes in reality reported after e.g. 14 months, and there are of course others (6 months, 24 months, 36 months). At least on the surface, it would seem adding support for specifying the sample period as X months back would be a relatively straightforward thing, and it would be very helpful for this purpose. I know people are using custom script for this type of thing today.
Thanks Nick, I think you describe a potentially «generic» approach to copying data N months forward which I didn’t think of! Essentially setting the sequential sample count to N and the skip count to N-1 (or N+1 and N - will have to test). If there aren’t any limits in how many months back you could go, this could take care of any cohort period, e.g. 12, 24, 36 etc.
I am not sure about all cohort types, would have to look a bit more. I believe for 6 months cohort it might work like this (remember I haven’t tested).
Period type = Monthly
Annual sample count = 0
Sequential sample count = 6
Sequential skip = 5
Then it would exclude the first 5 months of data in relation to the sample selected while also going 6 months back.
I guess you could potentially do the same for the other cohorts as long as your skip identified how far back you went?
I wanted to check if anyone had any experience using predictors for 1) «carry forward» of values from one period to the following, or 2) copying cohort data forward?
As far as I can tell from some limited testing, 1) should work as it is now by setting sequential sample count to 1 and annual sample count to 0. You could then for example copy ending balance for May 2017 to starting balance June 2017 etc. (note: I don’t mean to start a discussion how to collect data on stock).
Moving forward cohorts would similarly work today, but only for 12 month cohorts as far as I can figure out. From what I understand, even 12-month cohorts are sometimes in reality reported after e.g. 14 months, and there are of course others (6 months, 24 months, 36 months). At least on the surface, it would seem adding support for specifying the sample period as X months back would be a relatively straightforward thing, and it would be very helpful for this purpose. I know people are using custom script for this type of thing today.
Thanks, that’s neat. These are uses we never thought of for predictors!
There are no limits on how far back you can go with sequential and skip sample counts, so I think your technique for bringing forward cohorts should be good.
Cheers,
Jim
···
On Thu, Jun 29, 2017 at 8:07 AM, Olav Poppe olav.poppe@me.com wrote:
Thanks Nick, I think you describe a potentially «generic» approach to copying data N months forward which I didn’t think of! Essentially setting the sequential sample count to N and the skip count to N-1 (or N+1 and N - will have to test). If there aren’t any limits in how many months back you could go, this could take care of any cohort period, e.g. 12, 24, 36 etc.
I am not sure about all cohort types, would have to look a bit more. I believe for 6 months cohort it might work like this (remember I haven’t tested).
Period type = Monthly
Annual sample count = 0
Sequential sample count = 6
Sequential skip = 5
Then it would exclude the first 5 months of data in relation to the sample selected while also going 6 months back.
I guess you could potentially do the same for the other cohorts as long as your skip identified how far back you went?
I wanted to check if anyone had any experience using predictors for 1) «carry forward» of values from one period to the following, or 2) copying cohort data forward?
As far as I can tell from some limited testing, 1) should work as it is now by setting sequential sample count to 1 and annual sample count to 0. You could then for example copy ending balance for May 2017 to starting balance June 2017 etc. (note: I don’t mean to start a discussion how to collect data on stock).
Moving forward cohorts would similarly work today, but only for 12 month cohorts as far as I can figure out. From what I understand, even 12-month cohorts are sometimes in reality reported after e.g. 14 months, and there are of course others (6 months, 24 months, 36 months). At least on the surface, it would seem adding support for specifying the sample period as X months back would be a relatively straightforward thing, and it would be very helpful for this purpose. I know people are using custom script for this type of thing today.