New Contributor
Posts: 2

# Can it be predicted with time series?

Hi,

I have two dataset, one is rate_now, another is rate_lastyr. From rate_lastyr, you could see the rate is coming closer to a stable number after 1/22/2016, and ending up at 39.9% at month end. I want to predict what number rate_now will end up at based on the data from 1/1/2017 to 1/23/2017. Can I use time series? If so, how to predict? Thanks!

data rate_now;
input date \$ rate_now;
cards;
1/1/2017 45.6%
1/2/2017 47.1%
1/3/2017 47.1%
1/4/2017 45.1%
1/5/2017 41.9%
1/6/2017 41.1%
1/7/2017 40.9%
1/8/2017 40.9%
1/9/2017 40.8%
1/10/2017 41.1%
1/11/2017 41%
1/12/2017 40.8%
1/13/2017 40.6%
1/14/2017 40.4%
1/15/2017 40.4%
1/16/2017 40.6%
1/17/2017 40.7%
1/18/2017 40.7%
1/19/2017 40.3%
1/20/2017 40.2%
1/21/2017 40%
1/22/2017 40%
1/23/2017 40.1%
;
run;

data rate_lastyr;
input date \$ rate_lastyr;
cards;
1/1/2016 45%
1/2/2016 45%
1/3/2016 45%
1/4/2016 46%
1/5/2016 44.6%
1/6/2016 43.7%
1/7/2016 41.4%
1/8/2016 41.1%
1/9/2016 41%
1/10/2016 41%
1/11/2016 41.3%
1/12/2016 41.1%
1/13/2016 40.9%
1/14/2016 40.8%
1/15/2016 40.7%
1/16/2016 40.6%
1/17/2016 40.7%
1/18/2016 40.7%
1/19/2016 40.8%
1/20/2016 40.6%
1/21/2016 40.1%
1/22/2016 39.9%
1/23/2016 39.9%
1/24/2016 39.9%
1/25/2016 39.9%
1/26/2016 39.9%
1/27/2016 39.9%
1/28/2016 39.9%
1/29/2016 39.9%
1/30/2016 39.9%
1/31/2016 39.9%

;
run;

Posts: 1,122

## Re: Can it be predicted with time series?

This is not an answer to your question, but what rate is this?

Is the Jan 2016 data fixed at 39.9% as of the second-to-last Friday because it actually stabilizes, or is there some exogenous monthly event that causes it to be fixed near the end of each month.  I ask because there are some financial instruments that have a price or some other attribute fluctuate in a market most of the month until an authority (the "exogenous event") sets a fixed value near the end of the month, which then holds until the last day of the month.

New Contributor
Posts: 2

## Re: Can it be predicted with time series?

Hi Mkeintz,

Thank you for responding. The rate is the decline rate, which means, we have both an automatic deline system and human review process for declining the customers, for those who are sent to the human review process, we would spend some days finalizing the decision. So for the monthly decline rate, we have to wait till after mid-month, to get it stable. Thus, the Jan 2016 data was fixed at 39.9%  because it actually stabilizesfafter mid-month.

My problem is, how can I predict the stable decline rate for Jan2017 based on Jan1-Jan22,2017 and Jan 2016 in a more accurate way, so that I could predict Feb2017 based on Feb1-Feb22,2017 and Feb2016 next month and so is March and April and May...... in the same way?

Thank you so much!

Discussion stats
• 2 replies
• 115 views
• 0 likes
• 2 in conversation