Weekly and seasonal forecast

[#1]

Hi All

I have daily patient ED cases data for past 3 years. The initial trend run it is obvious that it has seasonality ( high in summer and  low in winter) .At the same time it also has weekly trend ( high on Monday)

How can we deal with two level of adjustment? I want to incorporate both ( seasonal and weekly) to suggest for optimal staffing  models.

Please advise. 

Thank you

[#2]

I suggest that you look into Time Series Analysis. Look into the JMP documentation on "Predictive and Specialized Modeling".
Help==>Books==>Predictive and Specialized Modeling

[#3]

Hi Nelson

Thank you. I read thru all of this and still couldnot figure out thw ay to admust multi sesonality in JMP. I am really struggling here. Please advise.

[#4]

Hi Chuiee -- what is it exactly that you're trying to do? Are you trying to find a way to visualize your data on both yearly and weekly time scales? To use both season and day-of-week in a predictive modeling approach? Or something else?

[#5]

hi
I have daily data. I need to forecast for an entire new year (365 days)
where it reflect the high in Monday and low in sundays throughout the year
and also the forecast should show high in winter and low in summer
simultaneously.

[#6]

I'm still not totally clear on what you want when you say "I need to forecast," but I'm going to guess you are looking for a compact way to visualize both the seasonal and weekly effects at the same time. If I were doing this, I might use two separate plots displayed next to each other, since it's different to see one effect that varies on a 7-day scale at the same time as one the varies on a ~180-day scale in a simple time trend unless it is very regular. Perhaps one would have the full year as the x axis, and the other would have each day of the week as a different bucket. Another way you could do it is with a variability plot that uses two levels of bucketing: weekday and season. I mocked up some similar data so I could show you this latter option. You can see right away that Winter is low and Summer is high, but within each bucket, Sunday (in my dataset) tends to be the highest day and it gets lower as the week goes on. Does this help at all?

[#7]

Hi Michele,

This is excellent.  I am very impressed. Will you please kindly show me how did you create this graph and forecast?

In addition to that, I want my 2020 FY ( july to june)  to look like this :

where it  shows the seasonal trend ( high in winter) like previous FY data ( 3 years) adn aslo  mondays high and sunday lows.

How do we do that?

[#8]

Hi There,
I am waiting on how did you developed this forecast and graph... Please reply.

[#9 · Accepted solution]

Oops, sorry I missed this!

I made my chart (in JMP 12) using Analyze > Quality and Process > Variability / Attribute Gauge Chart. I put Season followed by Weekday in the "X, Grouping" role and Forecast in the "Y, Response" role. Using the little red triangle menu you can add other things to the plot, like group means, mean diamonds, etc.

In the time trend plot you showed, I think you could illustrate the seasonal trend by legending by season -- it would become immediately apparent by color that one time of year is consistently high and one time of year is consistently low. However, I don't think you can show the annual periodicity in the same plot as the weekly periodicity, just because the weeks are too smushed-together (a technical term ;) ) to be visible on a year time scale.

[#10]

Thank you so much. You are awesome