# ISM PMI data - sector trend through ranking and seasonal decomposition

I have monthly data for each of the 18 sectors in the ISM PMI.

Each datapoint shows the trend of a sector: growing, contracting or neutral.
It also tells the strength of that trend with a number:
"Growing 1" is the sector that shows the mosts growth in that month.
"Contracting 1" is the sector that has the biggest contraction in that month.
Neutral always has the number 0.

So, for example, if you'd have to list a number of sectors from biggest growth to biggest contraction, it would be:
Sector A: Growth 1 (==> biggest growth)
Sector B: Growth 2
Sector C: Growth 3
Sector D: Growth 4
Sector E: Neutral 0
Sector F: Contraction 3
Sector G: Contraction 2
Sector H: Contraction 1 (==>biggest contraction)

My goal is to be able construct a chart like this,
which shows the underlying trend of a sector (dashed line):

The dashed line represents the underlying trend of the ISM "Computer & Electronic Products" sector. It's the ISM sector's ranking on a 1 to -1 scale and ran through seasonal decomposition (left hand scale).

The way I see it, it's a 3-part problem:
1. Ranking: For each month, the sectors need to be ranked (biggest growth to biggest contraction) and get a ranking number.
2. Scaling: Put the ranking number on a 1 to -1 scale
3. Seasonal decomposition: For each sector ("Computer & Electronic Products" in the example chart), you have a datapoint for each month. This constitutes a time series, which can be used for seasonal decomposition to discover the actual trend.

However, for each of these 3 steps, I don't how to perform the calculations.
Who can provide tips or examples on how to accomplish this please?

Example data
Should you need data to work with, I have put an excel with monthly data of all the ISM Manufacturing Sectors as of January 2008 here: https://goo.gl/IUuaj2

• Just a small comment from my side. This chart does not say anything AT ALL - and this is the most friendly way for me to put it. I don't see what people get of posts like "It might be clear, it might not be!" - this is a no-brainer. I would recommend anyone to be very careful with this "macro" charts with left and right scales and different origins (especially comparing leading and lagging numbers and not shifting). – vanguard2k Jun 23 '15 at 8:21
• But to deal with your question: What framework do you have? In the programming language R for example, you can use the command order() to rank the sectors, for seasonal decomposition you can do all kinds of stuff. As a first step I would use a moving average or the decompose() function of the stat package. As far as scaling is concerned, you will divide by the number of sectors at some point - but you can do it in different ways. The question is: do you want to preserve the information that the sector has a PMI that indicates growth or do you just want to see the ranking. – vanguard2k Jun 23 '15 at 8:21
• @vanguard2k Thank you for your quick reply. I don't really have a framework. I just use Excel. However, I am a programmer by profession, so I guess I could manage another software. But to keep it simple, I'd prefer to continue in Excel (if possible). I'm not a mathematician, but I have an engineering background, so I'm able to grasp math and statistics. To answer your question: I want to preserve the information that the sector indicates growth or contraction. The ranking will be needed to get there I think. I just don't know how. – Bjorn Mistiaen Jun 23 '15 at 17:59
• @vanguard2k My goal is to construct a graph that represents the base trend of the growth/contraction of that sector. I understand from your explanation that there are several ways to do the scaling, and also to do the seasonal decomposition. Could you point me to some resources or examples that explain these different ways, or some examples how to calculate these please? And also which would be the best method(s) for my case? – Bjorn Mistiaen Jun 23 '15 at 17:59