calculating averages with multiple zero values

[jimmyjump75]

Hi,

I am trying to accurately calculate average when there are many zero values. I have a computes. From this 10 computer I am collecting performance statistics for metrics of operations per second, bytes per second etc... I am collecting samples 3 times a minute or every 20 seconds for 3 days. So for this computer I have 180 samples an hour and 4320 samples per day. Every sample can range from 0 to 3000. I want to obtain the average per hour for each metric I am collecting statistics for.

The problem is that for many samples I have zero values and I am not sure how to properly calculate the average per hour. Usually I would say that for a specific computer for the hour of 11 AM to 12 PM I would add up the 180 samples and divide the sum of those samples by 180. However the zero values seem to skew the numbers to be really low. Is this the proper way to calculate average when you have many zero values?

Thanks
Jimmy

 

[Deleted member 4993]

Hi,

I am trying to accurately calculate average when there are many zero values. I have a computes. From this 10 computer I am collecting performance statistics for metrics of operations per second, bytes per second etc... I am collecting samples 3 times a minute or every 20 seconds for 3 days. So for this computer I have 180 samples an hour and 4320 samples per day. Every sample can range from 0 to 3000. I want to obtain the average per hour for each metric I am collecting statistics for.

The problem is that for many samples I have zero values and I am not sure how to properly calculate the average per hour. Usually I would say that for a specific computer for the hour of 11 AM to 12 PM I would add up the 180 samples and divide the sum of those samples by 180. However the zero values seem to skew the numbers to be really low. Is this the proper way to calculate average when you have many zero values?

Thanks
Jimmy

That will depend on whether those values are really zeroes or missing value.

If those are "missing values" - the need to be excluded and the count drops.

If those are really zeros then those need to be included in your calculation of average.

 

[JeffM]

Hi,

I am trying to accurately calculate average when there are many zero values. I have a computes. From this 10 computer I am collecting performance statistics for metrics of operations per second, bytes per second etc... I am collecting samples 3 times a minute or every 20 seconds for 3 days. So for this computer I have 180 samples an hour and 4320 samples per day. Every sample can range from 0 to 3000. I want to obtain the average per hour for each metric I am collecting statistics for.

The problem is that for many samples I have zero values and I am not sure how to properly calculate the average per hour. Usually I would say that for a specific computer for the hour of 11 AM to 12 PM I would add up the 180 samples and divide the sum of those samples by 180. However the zero values seem to skew the numbers to be really low. Is this the proper way to calculate average when you have many zero values?

Thanks
Jimmy

An "average" is simply a computation that is supposed to give a "typical" or "representative" value of a group of data. In some cases, the "typical" value is misleading; the data are so different that what is fundamentally important is the extent of their differences. In other cases, what is "typical" is not misleading, but the correct method of averaging must be used to show what is "typical." There are a number of ways to construct an average, and they do not give the same answer.

Three common methods of averaging.

The arithmetic mean: add up all the values and divide by the number of values.

The mode: what value occurs most often.

The median: what is the value such that half the values are higher and half the values are lower.

Without understanding a lot more about what you are doing and why, anything I say may be incorrect. But here is one thought. You could say:

The arithmetic mean of of all the observations _ is _ (which will be a very low number), but during those periods when the number of observations is greater than zero, the arithmetic mean of those values is _. That would be one way to summarize the fact that a lot of the time, nothing happens, but when something does happen, it happens a lot.

In other words, your job is to summarize a bunch of data in a meaningful and honest way. If it takes two numbers to do that, well then it takes two numbers to do that. Make sense?