EART60061 Chap.2 Measuring Variability and the Standard Deviation
Measuring Variability and the Standard Deviation
Measuring Variability starts from the problem that no measurement of a continuous variable is the true value, so its quality must be stated before it can mean anything. Using two years of student heights, the chapter moves from comparing the shapes of frequency distributions to a single number.
The range uses only two values; deviations from the mean cancel to zero; removing their sign and dividing by the number of values gives a fair, assumption-free measure of spread. Squaring the deviations instead, dividing by n minus 1 for a sample and taking the square root gives the standard deviation, the parameter that links spread to the normal distribution.
The chapter also covers the spreadsheet habits the exercises depend on: automatic calculation, the FREQUENCY array function, absolute references and the choice between sample and population functions. Throughout, the emphasis is on why each step is taken: the range ignores most of the data, signed deviations cancel, and the divisor standardises for sample size.
Understanding those reasons makes the later move from spread to probability straightforward.
What this chapter covers
- 01
Frequency distributions of two samples
- 02
Range and its weakness
- 03
Mean absolute deviation
- 04
Sample standard deviation and n minus 1
- 05
STDEV, STDEV.S and STDEVP in Excel
- 06
Standardising spread for datasets of different size
Worked example · free
Compare two measures of spread for conductivity readings
- 1Mean = 1,070 / 5 = 214 uS/cm; deviations are -2, 6, -9, 4 and 1, summing to zero.
- 1Mean absolute deviation = (2 + 6 + 9 + 4 + 1) / 5 = 22 / 5 = 4.4 uS/cm.
- 1Squared deviations are 4, 36, 81, 16 and 1, totalling 138.
- 2Divide by n minus 1 = 4 to get 34.5, then take the square root: s = 5.87 uS/cm.
Key terms
- Frequency distribution
- A count of how many measurements fall into each class, used to compare the shape and spread of datasets.
- Mean absolute deviation
- The average distance of values from their mean, ignoring sign, which makes no assumption about how errors are distributed.
- Absolute reference
- A spreadsheet cell reference fixed with dollar signs so that it does not shift when a formula is copied.
- Population standard deviation
- The spread of a complete population, dividing by n rather than n minus 1, returned by STDEVP in Excel.
Measuring Variability and the Standard Deviation FAQ
Why can a measurement never be the true value?
Any measurement of a continuous variable carries error from the instrument, the procedure and natural variation. The module therefore treats a result as meaningful only when it is quoted with an indication of its quality, which requires repeated measurements so that the variability can be quantified.
Why divide by n minus 1 in the standard deviation?
The values are usually a sample of a larger population, such as 500 students out of a whole university. Dividing by n minus 1 instead of n corrects for estimating the mean from the same sample, and it is what STDEV and STDEV.S do in Excel.
What is the FREQUENCY function used for?
It sorts a dataset into classes for a frequency distribution. It is an array function: highlight the output cells, enter the data range and the class limits as arguments in the formula bar, then press Control, Shift and Enter together so the counts fill every cell at once.
Should I use the range or the standard deviation to compare datasets?
The range is quick but uses only the largest and smallest values, so one unusual reading can change it completely. The standard deviation uses every value and connects to the normal distribution, which is why the module builds towards it and uses it for confidence limits.
Why does the standard deviation square the deviations?
Squaring removes the sign so deviations no longer cancel, and it gives the parameter that controls the width of the normal distribution. The sign-free mean deviation also removes the sign but makes no assumption about how errors are distributed, so it cannot be turned into a probability in the same way.
Exam move
Do the long calculation once on a small dataset with a calculator, then reproduce it in Excel with deviation, squared deviation and sum columns before checking against STDEV. Write down which divisor each function uses. Then open the heights workbook, confirm that calculation is automatic and practise an absolute reference by copying a deviation formula down a column and checking that it still points at the mean.
Once the calculation is secure, compare the two years of heights with both measures and explain in one sentence why they rank the years the same way.