Wednesday, October 3, 2012

W8_TRI_ Billet Steel in Gold Equivalent Value

1.      Problem recognition, definition and evaluation

Steel is primarily used for major material in oil and gas projects. The volatility of steel price has therefore impacted actual projects cost overruns. Figure 1 shows the steel price has shown significant volatility for the last 5 years. Hence, the future cost estimation has to consider this positive trend of steel price.
Figure 1. World Average Steel Prices

2.      Development of the feasible alternative

This will explore gold as alternative reference currency to be used in steel price forecast particularly in pipeline projects that utilize billet and scrap steel. This chapter will develop model for billet price.

3.      Development of the outcome

The objective is to convert and use billet volume (tonne) into gold equivalent (ounce), so that billet price will refer to gold price. The historical data of monthly average price of billet steel in gold equivalent (tonne Billet/oz Gold) from Jan 1996 to July 2012 is presented in Figure 2.  
Figure 2. Monthly Average Steel Prices

4.      Criteria of Selection

By applying a simple linear regression in Figure 2, the billet-gold equivalent was steadily increasing as a function of time. However, S-squared of regression was very small that means it didn’t closely fit to the data. Therefore, control chart analysis will be used as benchmark in selecting the best method that represents the data. The parameters of control chart are derived as presented in Table 1.
Table 1. Billet in Gold Equivalent (96-12)

5.      Analysis

The control chart was built according to Table 1 as shown in Figure 3. 
Figure 3. Control Chart Billet-Gold Equivalent (96-12)
The chart shows that there is a new pattern of data after January 2009 due to extremely increament of gold price . And thenTable 2 and  Figure 4 pointed that period out in a new control chart to adopt new pattern in billet-gold equivalent.  
Table 1. Billet in Gold Equivalent (09-12)

Figure 4. Control Chart Billet-Gold Equivalent (09-12)
From Figure 4 shows that there was an outlier since it was out of contol limit (UCL/LCL), thus the control chart wasn’t valid referring to Shewhart X-charts rules. On the other side, this also shows that the polynomial regresion hasn’t  produced a minimum 70% R-Squared, in which for instance the 4th poly only produced 0.4146 R-squared.

6.      Selection of alternative

 Both methods (regression and control chart) actually didn’t fit the technical requirement to be chosen as model of billet price forecast using gold equivalent. However in my opinion, the control chart is better in term of data matching and more comfortable to be used in forecast purpose since it allows plus/minus sigma as a price range.

7.      Performance monitoring and post-evaluation of results

The control chart will be therefore used as reference chart in linking billet price in gold equivalent. The updated chart shall be applied to check the chart validity and improve its reliability.

References:
·         Brassard, M and Ritter, D (2010), The Memory Jogger (2nd Edition),Canada, GOAL/QPC
·         Ming, T. (2001, Feb). An Introduction to SPC: SPC Tools - Control charts. Retrieved from: http://lorien.ncl.ac.uk/ming/spc/spc8.htm
·         Steel on the net. (2012, Sep 24). Historical Steel Price. Retrieved from: http://www.steelonthenet.com/pricing_history.php
·         World Gold Council. (2012, Sep 24). Interactive gold price chart and downloads. Retrieved from: http://www.gold.org/download/value/stats/statistics/xls/gold_prices.xls


6 comments:

  1. Nice work, as usual, Pak Trian......

    Couple of suggestions...... To help get your R2 value up, try shortening up the time frame. Instead of working from 1996 to 2012, try analyzing only the past 100 months. (8.33 years) And if that doesn't work, then try the past 5 years (60 months)

    Another suggestion...... For Figure 1, try adding in the cost of an ounce of gold...... As an ounce of gold has historically been pretty close to a ton of steel, it will be interesting to see when the two started to diverge and have you explain why that is happening.....

    Bottom line, great topic, good start on your paper but there is a bit more you can do to "set the stage" for your analysis.

    BR,
    Dr. PDG, Singapore

    ReplyDelete
  2. Dr. PDG,

    I just figured out the highest R2 value was around 0.75 using 127 data after many times trial. I would say that's the best equation it can produce by considering 36 forward data (next 3 years).

    Similar problem i have faced for other commodities in which their R2 values were not too high (< 0.9).

    please give your suggestions..
    thanks

    ReplyDelete
  3. Have you solved this problem yet? Now that you have a range, does the price of steel in ounces of gold fall within that range?

    If not, then you conclude in your paper that using gold equivalence is NOT reliable for steel prices.......

    BR,
    Dr. PDG, Jakarta

    ReplyDelete
  4. After thinking on this a bit more, you don't have to use regression analysis on the ounces of gold per tonne of steel.

    All you have to do is apply statistical process control charts analysis to see if the ounces of gold per tonne of steel fall within +/- 3 sigma....

    If that is true, then you can use the data. If not, then you have to throw out the outliers.....

    I also suspect that if you took the same approach and looked at the interest rate vs ounces of gold per $1,000 that MAY solve your problem using that data as well.....??? Give it a try and see what you find....

    BR,
    Dr. PDG, Jakarta

    ReplyDelete
  5. Yes..Dr Paul.
    I resolved this problem by using what you suggested, taking off the outliers.

    I nearly finish my papers, i hope i can send a complete paper tonight. if it's not completed yet, i will keep sending it with some un-finished explanations (i have finished the calculation, tables, curves).

    Thanks

    ReplyDelete
  6. Looking forward to seeing your paper....!!!!

    BR,
    Dr. PDG, Jakarta

    ReplyDelete