Professional Documents
Culture Documents
Theweeklypricesacrossthedifferentregionsarehighlycorrelated:
MakingSenseofDieselPrices 1 SpiderFinancialCorp,2013
Process
Nowwearereadytoconductourprincipalcomponentanalysis.First,selectanemptycellinyour worksheetwhereyouwishtheoutputtobegenerated,thenlocateandclickontheprincipal component(PCA)iconintheNumXLtab(ortoolbar).
Theprincipalcomponentanalysiswizardappears.
MakingSenseofDieselPrices
SpiderFinancialCorp,2013
MakingSenseofDieselPrices
SpiderFinancialCorp,2013
Now,clicktheMissingValuestab.
MakingSenseofDieselPrices
SpiderFinancialCorp,2013
Analysis
1. Statistics
MakingSenseofDieselPrices
SpiderFinancialCorp,2013
2. Loadings
Intheloadingstable,weoutlinetheweightsoftheprincipalcomponentineachinputvariable:
Thisfactorsloadingcanbeviewedasproximityto(oravailabilityof)refinerycapacityorimportports (example:NewYorkharbor).Thesecondfactorreflectsthecostoftransportationandtaxofthefuel.
MakingSenseofDieselPrices 6 SpiderFinancialCorp,2013
MakingSenseofDieselPrices
SpiderFinancialCorp,2013
Thetwotimeseriesexhibitsomeseasonality,althoughitismoreapparentinthesecondfactor. Furthermore,thefirstprincipalcomponentexhibitsapatternclosetocrudeoilprices.
NOTE:ThelackofanexactmatchmaybeattributedtoothercostsincurredinthemakingofNo.2Ultra LowSulfurDiesel:labor,energyprices,rawmaterial,etc.Furthermore,refineriesbuildupinventoryof products(e.g.diesel)inanticipationoftheseasonaldemandpeaks,sotheremaybealag. 4. Adding WTI Spot Prices AsalastthoughtontheWTIspotprice,wewillincludetheWTIspotpriceinourinputdatasetandre examinetheinputloadings.Intuitively,addinganinputvariableforrawmaterialprice(i.e.crudeoil) withthefinishedproductspricesinthesamedatasetwilllikelyrevealanotherdriver:costof production.
MakingSenseofDieselPrices 8 SpiderFinancialCorp,2013
Now,wedneedthreedriverstoaccountfor99.8%ofthepricevariation.
Notes: 1. Theloadingsforthefirstcomponentaresimilartooneswecalculatedearlierwithonlythe dieselprices.NotethattheWTIloadingsareslightlylowerthantheirdieselcounterparts.Again, welldesignatethisfactorasthegeneralpricelevel(regionneutral). 2. Theloadingsforthesecondfactorareverydifferentnow,andtheloadingforWTIisnegative( %94.5)whilealltherestarepositive.Wecandesignatethisfactorasthecostofcrackingdiesel ofcrude. 3. Theloadingforthe3rdfactorisverysimilartotheloadingofthesecondcomponentinthe earlierdatasets.Again,theloadingvariesbasedonthelocation(eastvs.west).TheWTI,Gulf Coast,andMidwestarealmostneutral. Letsnowplotthefactorstimeseries.
MakingSenseofDieselPrices
SpiderFinancialCorp,2013
MakingSenseofDieselPrices
10
SpiderFinancialCorp,2013
Conclusion
Inthistutorial,weexaminedanontrivialapplicationforPCA.Weattemptedtoexplaintheprice variationamongdifferentPADDregionsbyuncoveringthedrivingforcesbehindtheprices. Whatsnext? Wementionedearlierthatrefineriesbuildupinventoryofproducts(e.g.diesel)inanticipationofthe seasonaldemandpeaks,sotheremaybealag.Howdowecapturethisvariable?Futures:wecan includethefuturepricesofthedieselandcrudeoil. Why? Intuitively,futurespricesreflectmarketanticipationof(1)futuredemand(2)futurestoragecost,and possiblyapremiumforsupplyscarcity. ThisapplicationisintendedtogiveyouasampleofhowtoapplyPCAandtimeseries,aswellashowto useorinterpretthevariablesloadingsinderivingapracticalproxyforthem. Insum,PCAisamathematicalprocedurethatNumXLcanhelpyouexecute.Makingsenseofand interpretingtheresultsiswhereyourexpertiseandintuitionareindispensable.
MakingSenseofDieselPrices
11
SpiderFinancialCorp,2013