# Percentage in Statistics

Hi there, me again…

Up here in the Scandinavian countries the system specifications is still written as an X number in X percentage of the seating area. Like “minimum STI 0,55 in 90% of the seating area” or ±3dB in each 1/3 octave band In 90% of the seating area”

A very nice and useful feature would be a second choice of statistics calculations. Today we have stdv., but if we could switch by a right click in the statistic window and choose “percentage” that would solve allot of work, and make modeler output more suitable to some system specifications.

Target frequency should be selected as usual in the toolbar “Frequency”, and a easy way to select an target STI would be to use the same toolbar but when “Speech” Tab is selected the “frequency toolbar change from frequencies to STI, PB or % ALcons, depending what is selected in Simulation>Data.

Best, jorgen

Hi Jorgen,

It is on the list for the next release to do a graph that has STI on the horiztonal axis and “accumulated frequency” on the vertical axis. With this graph you would be able to see / read that for example 53% of the audience have an STI above .55, or that 95% of the audience is above .52. I think this will solve the problem, correct?

And here I’m faking nice pictures and so on, when you guys have already done it…

Thank you so much!

It will be very appreciated!

Best, jorgen

Hi Jorgen,

As Morten mentioned this is something that we have added to the list for an upcoming release. The following is an example of how a % of mapped area (i.e. seating) vs. STI score might look.

The STI scale matches the color scale that we currently use for the STI map.

I also added an indicator bar - which might be user definable - to mark the > x% range.

What do you think of something like this?

I like this Rob,

Best,

Tom

Posted by RK for JA:

Hi Rob.
Your image is a little small but it looks great.

Question, your cursor displays 80% and are positioned at STI 0,25. Does this mean that minimum STI is 0,25 in 80% of the seating area?

And if you would move the cursor to the right the percentage will be less and STI higher?

If that’s how it works, then it will be very useful for us!

Is this new statistic replacing mean&stdv or as a second choice?

And does it also work in D and D R?

Still playing around with this idea, but here is an example of what a display like this could look like for a coverage map. Ideally you would be able to set the “bin” size for the chart. This image shows the same data with two diffrent bin sizes (ignore the percentages this is for look only).

I also think the user should be able to define the % bar to their liking.

rk

[ThomasS: Posted for JA, currently enjoying vacation at the other end of the planet:]

So if I read your diagram right, 80% of the area is within 0 to -4dB?

Maybe its already implemented, but it would be very useful if the samples on the map is grayed out below (to the left) the cursor. By doing so the user/consultant can easy identify were the area under the specification are.

Best, jorgen

Although we might want to know what percentage of the seats exceed .5 STI or what STI is exceeded in 80% of the seats, the same is not quite true for SPL. In that case we may want to know what percentage are /-2dB of the ideal spec. Is there a general way to deal with both “single-sided” and “double-sided” percentage questions? Yes.

A very common approach that applies to both cases (and others) is to allow the user to choose between graphing the frequency distribution {the statistic which we show now) or graphing the “cumulative frequency distribution” which has the same x-axis, but always has a y-axis going from 0% to 100%. An example is shown here. If you search on cumulative distribution function, there are many other examples that clarify the concept.

The nice thing about the CDF is that ALL questions about percentages are answered at once by one graph. No mousing required. You can see what percentage is above, below, within, or outside some range of values of STI or dBSPL. Or inversely, you can specify a particular percentage and find the exact value of STI or dBSPL that is exceeded or not met in that percentage of the seats. It works for /- limits as well as single-sided requirements. And we might find other places in Modeler where the CDF technique would be useful. Basically any statistical graph showing a histogram of frequency of occurence of anything is more useful if one can also see the corresponding CDF graph, either superimposed or as an alternative.

Hope this is a helpful suggestion.
Chris