## Controller (SaaS, On Premise)

cancel
Showing results for
Did you mean:

Voyager

## Calculations for standard deviations

Hi all,

I want to exactly understand how dynamic baselines are calculated.

Let's assume I have the observations v_1,..,v_n

```std dev = sqrt [(B - A^2/N)/N]
where
A is the sum of the data values.
B is the sum of the squared data values.
N is the number of data values.```

How to determine if  v_{n+1} is within 3 times the baseline?

Is the following true?

v_{n+1} is within 3 times the baseline if and only if

v_{n+1} < 3*(std dev)

Are they any documented example about standard deviations / moving average / exponential smoothing within AppDynamics?

Best,

Daniel

6 REPLIES 6
Architect

## Re: Calculations for standard deviations

Hi Daniel, the following does look like the case it is done.

v_{n+1} is within 3 times the baseline if and only if

v_{n+1} < 3*(std dev)

AppDynamics Team

## Re: Calculations for standard deviations

Hi Daniel

To add on example on this,
For example, if BV = 100 and if condition is set as alert if observed/evaluated value is 2 StdDev below BV then:

A) StdDev = 10
Observed/Evaluated value < 100 - (2 * 10) ==> 100 - 20 ==> 80
Which means below 80, the alert will be fired.

B) StdDev = 2
Observed/Evaluated value < 100 - (2 * 2) ==> 100 - 4 ==> 96
Which means below 96, the alert will be fired.

To get a better view of what the standard deviation is for any metric, if you graph it in a metric browser and select the required baseline + select the baseline checkbox, then hovering mouse over the dash line graph will show you the StdDev and baseline observed value.

Thanks,

Yogesh

Found something helpful? Click the Accept as Solution button to help others find answers faster.
Liked something? Click the Thumbs Up button.
Voyager

## Re: Calculations for standard deviations

Hi yogesh,

does BV stands for baseline value? How does a fixed baseline comes along with the dynamic baseline?

From my understanding you only need the following values to determine if your current value is within k*stddev

Input

Current value v_{n+1}

Values v_{1},..,v_{n+2}

from the values you can just calculate the std dev from the above listed equation. I does not understand where your BV comes from? It should actually not predefined but calculated with the above listed equation?

Looking forward for a more precise understanding of your example! :-)

Best,

Daniel

AppDynamics Team

## Re: Calculations for standard deviations

Hi

Yes, BV is a baseline value. There are built-in baselines as well as you can configure it as per your requirements from the Configuration > Baselines page. You can check the baseline value for a particular metric in the metric browser.

I would request you to go through the below documentation for more information.

https://docs.appdynamics.com/display/PRO44/Dynamic+Baselines#DynamicBaselines-ViewBaselines

Thanks,

Yogesh

Found something helpful? Click the Accept as Solution button to help others find answers faster.
Liked something? Click the Thumbs Up button.
Voyager

## Re: Calculations for standard deviations

Hi yogesh,

I think my misunderstanding is coming from the following:

1) I assumed that your BV is a pre defined constant

2) Actually the BV is actually calculated on the past baseline (like Weekly Trend - Last 3 months)

From my understanding the Baselinen value is actually the standard dev. How could I calculate the Baseline Value of the following random values?

 sum(v_1,..,v_7)= sum(v_1²,..,v_7²)= 17,19919388 57,63445513 A B Parameter value v_1 4,508075756 20,32274702 N 7 v_2 2,945633959 8,67675942 std dev(v_1,..,v_7) 1,482062242 v_3 4,231876722 17,90878059 v_4 0,292167226 0,085361688 v_5 1,539746529 2,370819373 v_6 0,97697392 0,954478041 v_7 2,704719765 7,315509007 v_8 4,859193049

std_dev could easy be calculated through

$std\_dev = \sqrt(\frac{B - \frac{A^2}{N}}{N})$

What is the equation for the baseline value? From my initial thinking is it equalivalent to std_dev.

Can you please confirm the following definitions in the context of the above listed table and equation?

the "baseline" is a vector A = (v_1,v_2,v_3,v_4,v_5,v_6,v_7)

vector B = (v_1²,v_2²,v_3²,v_4²,v_5²,v_6²,v_7²)

N = 7

the "baseline deviation" is = std_dev (A,B,N)

Voyager

## Re: Calculations for standard deviations

Any idea on this?