Download as pptx, pdf, or txt
Download as pptx, pdf, or txt
You are on page 1of 19

Normal distribution

Standard normal distribution


MPH

Dr Helmy Hazmi
The shape that we would get when we graph the population frequency for continuous data
The normal distribution curve
Theoretical concept only

We can only use samples to estimate such data

Assuming that the samples fit the normal curve, we can make statistical
estimates about the population parameters
Gaussian distribution
Carl Friedrich Gauss
(1777-1855)

Properties of the normal curve


Smooth. Symmetrical. Bell shaped
Most observation clusters around mean
Mean = median = mode
50% of the area under curve is to the right of a perpendicular erected at mean
68.95.99.7
µ shift the curve on the x-axis
σ determines the degree of flatness or peakness of curve
Observation decreases steadily in opposing directions

No exact boundaries to the curve, so the curve never touch the baseline
The mean (SD) weight in a sample population is 65.1 (6.06)

In the population, what does that mean?


34.13% of area under
the curve is bounded
by the mean and the
score one SD above or
below the mean
Same mean but different SD

The bigger the SD, the flatter the


curve.
The Standardised Score
Since the units in scores are diverse, we need to have a standardise
method to make statistical data more meaningful and easier to
understand.

One way is to express scores in terms of standard deviation.

This is called the z- scores.


𝑋−𝑋
𝑧=
The formula: 𝑆𝐷
Using the previous example, what is the
z-score
of a person who weighs
75.4 kg?

Show the diagram to aid explanation.

Extra:
What is the proportion of our sample who are above 75.4kg?

You’ll need to refer to Area under the normal Curve table (z)
Skewness
Practically, not all data can fit in the normal distribution curve - especially small
samples.
Is a lack of symmetry of a distribution curve.

Reasons
to need to know shape of data:

Most statistical analysis are based on assumptions of normal distribution


Alternative statistical analysis are available for skewed data
Alternative procedures can transform skewed data

INVALID ANALYSIS
Negatively skewed = skewed to the left
Mode > median > mean
Median – quartile 1 > quartile 3 - median

Positively skewed = skewed to the right


Mean > median > mode
Quartile 3 – median > median – Quartile 1
Measure of skewness
Pearson coefficient of skewness

Positive value: Positive skew


Negative value: negative skew
Measure of kurtosis
Relative concentration of scores in the center, the tails and the
shoulders of a distribution
The variation in symmetrical distribution
The peakness
Terms (historical purpose):
- leptokurtic: higher peak than normal curve
- Platykurtic: the peak is more flat top than normal curve
- Mesokurtic: normal curve
Determine normality in SPSS
1. Analyze > descriptive statistics > descriptive > option > check kurtosis and
skewness > enter
2. Analyze > descriptive statistics > explore > put in variables in the dependent
list > enter

Skewness
The ratio of skewness statistic to the error should not be more than -1.96 to +1.96

Kurtosis
The ratio of kurtosis statistic to the error should not be more than -1.96 to +1.96
Determine normality in SPSS
Analyze > descriptive statistics > explore > put in variables in the
dependent list > enter

Kolmogorov smirnov test


Used when n>2000

Shapiro wilk test


- Used when n <2000
Read here.

N=20 (ie use Shapiro wilk)


P=0.316 (ie >0.05)
Therefore we accept the null hypothesis,
normality is assumed.

You might also like