Question around Correlation

Hi,

Quick question around Correlation. User Manual 5.2.1 on page-54 says that

“There are many ways to correlate data using various statistical methods. PMF uses the

percent achieved of an objective or measure to determine how points of data are plotted on the Correlation scatter plot.”

Can correlations be done using 'actual' figures and not the "percentage achieved" for the measure? I think it’s a FOCUS report which runs on background and by default setting looks like “percentage achieved” instead of actual. Can we change it somewhere in code so that our user can correlate actual figures?

In analysis designer mode, how about if we have correlation feature. Not sure if its a good idea?

Thanks

Abhinav

Bob Jude Ferrante

Master

posted

Hide Post

The reason Percent Reached rather than actual value was chosen was threefold:

The raw value of a metric eliminates all notion of its goal. For a sales measure expressed in currency, you might have one office making $1000.00 and another making $100,000.00; now you might think the second office was doing much better than the first, until you realized the first office only had a goal to make $1100.00 so they're actually doing quite well, whereas the second office had a goal of $1,000,000.00 which means they're doing quite poorly.
There's this notion of the direction of a measure, which is embedded in Pct Reached but not present when you're using an Actual value. So an increasing value isn't always a good thing if you're talking about measures like Injuries on a plant floor, Employees Resigning in HR, or Fatalties in a hospital. Correlation is based on assuming the direction of a number is the same for all examples. So correlating a descending Measure like Injuries to an ascending Measure like Profit would give incorrect correlative results. With Pct_Reached you have the notion of direction built in.
Some measures can vary greatly in magnitude - you could have a measure in *Hundredths* and another in *Trillions*, and the Hundredths would be totally invisible - far less than a "rounding error" on the X or Y axis if the other measure was in a different magniture. With Pct_Reached, the target is scaled to the Actual, so you end up comparing things that show up.
You can store percents in measures in which case the Actual = Actual_Numerator/Actual_Denominator. Pct_reached deals with this well.

Now, if you can reassure yourself that neither you nor your stakeholders would try to correlate with any Measures that would have any of the inclusions above, you could change correlation to print the RPT_ACTUAL column in the FOCEXEC that correlate uses (which is correlate_hold.fex). Make sure you alias the fields you output using AS names so the further down the chain FOCEXECs (correlate_table.fex and correlate_gph.fex) see the fields they expect so see. Put your modified FOCEXEC in the pmfcustom app folder and make sure that folder appears in the WebFOCUS APP PATH before the mainstreet folder, as always (you should never put mods into the core folders because they get overwritten on upgrades; we promise never to touch your pmfcustom folder).

Long-term, possibly we could consider at least trying to show the equivalent Actuals next to their Pcts Reached on the grid portion of the output, and perhaps some way to display them in the hover text on the chart.

thanks for the suggestion.

Bob Jude Ferrante
Director of Business and Development
WebFOCUS Performance Management
Bob_Ferrante@ibi.com
917-339-5105

I'll take any questions about PMF - business or technical - anytime!

Posts: 919 | Registered: March 26, 2003

Please Wait. Your request is being processed...

Read-Only Topic

Focal Point

Focal Point Forums

Performance Management Framework (PMF)

Question around Correlation

Focal Point | Privacy Statement | Terms of Service