Key Influencer - likelihood calculation not matching

[Anonymous]

HI there,

I've created this Key Influencers visual but don't figure out the calculation for the likelihood. 

The green bar is at 1.35% (2747)

Average is 0.78% (36548)

The likelihood plotted is at 1.82x.  it's not 1.35 / 0.78...

There are no blanks or nulls in the column feeding the Explain by filed...

What am I missing ?

Thank you

[Anonymous]

Hello all,

After some more digging, I understood that the visual takes a sample of 10k rows to build the analysis on. So if you load less than 10k rows, the division should work out, if more than 10k in the data, the sampling (stratified sampling) will grab the most representative 10k only and this is why the division will not match.

Thanks all,

Mihnea

[Anonymous]

---

@Anonymous wrote:

Hello all,

After some more digging, I understood that the visual takes a sample of 10k rows to build the analysis on. So if you load less than 10k rows, the division should work out, if more than 10k in the data, the sampling (stratified sampling) will grab the most representative 10k only and this is why the division will not match.

Thanks all,

Mihnea

---

Hi,

How did you learn about that? Can you please recommend any article explaining how stratified sampling is performed in Power BI? 🙂

[Olimpiusz]

Och! Thank you for pushing me in this direction. How did you notice that? Any Microsoft article or just an assumption?

[PJ-292]

were you able to solve this? i'm having the same trouble of the likelihood figure not matching up too. 

[Olimpiusz]

still waiting for answers.. how can we use it well if we do not know the basics.. 😉

[Olimpiusz]

Any news on that one? I'm also very interested in this topic. Thanks!

[dax]

Hi mikyi_ro,

The likelihood seems not equal to green bar /red line(it approximately equal) , it seems that it has its algorithm. You alo could refer to Key influencers visualization(you will see the likelihood  also not equal to gren bar /red line). I will discuss with other engineers about this algorithm, and I will inform you as soon as i get it.

Best Regards,
Zoe Zhi

[Olimpiusz]

On Misrosoft page you wrote that:

So how we should interprete that?

[Anonymous]

OK, Dax, thank you,

will wait for the feedback from the engineers

thanks