Finding a Missing Value

[KLL]

Hello,

I have data that I'm trying to trend and some values are missing so I've been doing a simple calculation to find an estimate of what the missing value is. I just want to know if my thinking is correct.

Every month I collect an operations counter from some equipment, but some months I might miss it.

For example:
Jan = 654104
Feb = Missing
Mar = 659176

So, I take 659176 - 654104 = 11359, then I divide 11359/2 = 5679 I then add 5679 to Jan 654104 = 659783

Now it would look like
Jan = 654104
Feb = 659783
Mar = 659176

Is this correct?

 

[Dr.Peterson]

I have data that I'm trying to trend and some values are missing so I've been doing a simple calculation to find an estimate of what the missing value is. I just want to know if my thinking is correct.

Every month I collect an operations counter from some equipment, but some months I might miss it.

For example:
Jan = 654104
Feb = Missing
Mar = 659176

So, I take 659176 - 654104 = 11359, then I divide 11359/2 = 5679 I then add 5679 to Jan 654104 = 659783

Now it would look like
Jan = 654104
Feb = 659783
Mar = 659176

Is this correct?

It would be a valid method (for linear interpolation), if you had done the arithmetic correctly!

When you subtract 659176 - 654104, you should get 5072; I don't know how you got 11359.

And in the end, did you look at the number you got, expecting it to be between the two known numbers? That's a simple "sanity check". It should have been obvious that your answer was wrong, since 659783 is greater than either.

Also, there is an easier way in this case (though what you did is excellent if you hadn't wanted the middle value). You could just average the two outside numbers by adding them and dividing by 2. That would give the same result as your calculation, done correctly.

There are other potential issues that others might want to deal with, but we'd need more data to see whether averaging is the best thing to do.

 

[blamocur]

Hello,

I have data that I'm trying to trend and some values are missing so I've been doing a simple calculation to find an estimate of what the missing value is. I just want to know if my thinking is correct.

Every month I collect an operations counter from some equipment, but some months I might miss it.

For example:
Jan = 654104
Feb = Missing
Mar = 659176

So, I take 659176 - 654104 = 11359, then I divide 11359/2 = 5679 I then add 5679 to Jan 654104 = 659783

Now it would look like
Jan = 654104
Feb = 659783
Mar = 659176

Is this correct?

In addition to @Dr.Peterson's comment I'd mention that if those numbers accumulate over time you might want to remember that February is several days shorter than March or January.

 

[KLL]

Sorry about that I am working with a ton of data and didn’t catch that mistake.

For example:
Jan = 654104
Feb = Missing
Mar = 659176

So, I take 659176 - 654104= 5072, then I divide 5072/2 = 2536 I then add 2536 to Jan 654104 = 656640

Now it would look like
Jan = 654104
Feb = 656640
Mar = 659176

Now if I do it the way you mentioned above by adding 654104 + 659176 = 1313280 and dividing 1313280/2 = 656640
It’s the same answer what’s the difference between that way and mine?

Basically I do monthly inspections of equipment. I gather the operations counter each month. Some months I can’t due to other things that might pop up.

Counter
Jan = 654104
Feb = missing (populate with calculated 656640)
Mar = 659176
April= 661476
May= 664376
June= missing
July = missing
August = 671976
September= 675076

I then find the total changes month to month

Total Changes
Jan - Feb = 2536
Feb- March = 2536
March-April= 2300
April-May= 2900

After I have a years worth of data I sum the total changes. To see how much it operated in a year. So above from Jan - May would be a total of 10,272 operations.