Parametric Statistics

download report

Transcript Parametric Statistics

Parametric Statistics
Descriptive statistics
Hypothesis testing
• Population: the entire set about which info is
needed, Greek letters are used
• Sample: a subset studied; random samples
• Parameter – numerical characteristic
• Inferential statistics
• Discrete and continuous
• Distribution: the pattern of variation of a variable
• One sided probability: comparing two data sets
(ie. a > b); two-sided probability: a not equal to b
Detection Limit
Action Limit; 2s, 97.7% certain that signal observed is
not random noise.
Detection Limit; 3s, 93.3% certain to detect signal
above the 2s limit when the analyte is at this
• Quantitation Limit; 10s signal required for 10% RSD
• Type I Error: identification of random noise as signal
• Type II Error: not identifying signal that is present.
Numerical Descriptive Statistics
Types of numerical summary statistics:
• Measures of Location
Measures of Center
Other Measures
Measures of Variability
Probability Density Function
• Probability density function – f(x) –
probability of obtaining result x for the
variable in your experiment (relative
frequency for discreet measurements)
• The total Sf(xi)=1 the total sum of all relative
• Distribution function – probability that x is
less than or equal to xi:
– F(xi)= = Sf(xj) over all j such that xj xi
Discreet Data PDFs
• Binomial distribution:
f(x) = (n!/(x!(n-x)!))px(1-p)n-x
• The probability of getting the result of
interest (success) x times out of n, if the
overall probability of the result is p
• Note that here, x is a discrete variable
– Integer values only
Uses of the Binomial Distribution
• Quality assurance
• Genetics
• Experimental design
Binomial PDF: Example
n = 6; number of dice rolled
pi = 1/6; probability of rolling a 2 on any die
x = [0 1 2 3 4 5 6]; sample results # of 2s out of 6
Graph of f(x) versus x for rolling a “2”
Binomial PDF: Example 2
• n = 8; number of puppies in litter
• pi = 1/2; probability of any pup being male
• x = [0 1 2 3 4 5 6 7 8] example data for the # of
males out of 8
• Graph of f(x) versus x
Binomial PDF: Characteristics
• Shape is determined by values of n and p
(parameters of the distribution function)
– Only truly symmetric if p = 0.5
– Approaches Poisson’s distribution if n is very
large and p is very small,
– Approaches the normal distribution if n is large,
and p is not small
• Mean number of “successes” or the
expectation value X = np
• Variance is np(1- p)
Poisson Distribution
• Can be derived as a limiting form of Binomial
– when n∞ as the mean l=np remains constant
– this means conducting a large number of trials with
p very small
• Can be derived directly from basic assumptions
• Assumptions determine the real situations where
Poisson’s distribution is useful
Simeon D. Poisson (1781-1840)
Poisson’s Assumptions
• Time or other interval type study
• The time interval is small
• The probability of one success is proportional
to the time interval
• The number of successes in one time interval
is independent of the number of successes in
another time interval
Derive Poisson from basic
• Derivation by Induction
– To find an expression for p(x), first find p(0),
then p(k) and p(k+1) then generalize for p(x).
• Basic properties used:
Poisson’s Assumptions: Example
• The probability of one photon arriving in
the time interval Dt is proportional to Dt
when Dt is small
• The probability that more than one photon
arrives in Dt is negligible for small Dt
• The number of photons that arrive in one
time interval is independent of the number
of photons that arrive in any other nonoverlatping interval
Normal Approximation to Poisson’s
Measures of Center
• Mode
• Median
• Population Mean (μ) and Sample Average (x)
Measures of Spread
• Variance – square of standard
• Standard deviation:
– Population standard deviation s:
large sample sets, the population
mean (μ) is known.
– Sample standard deviation (s):
small sample sets, sample average
(x) is used.
– Pooled standard deviation (s ).
When several small sets have the
same sources of indeterminate
error (ie: the same type of
measurement but different
Standard Error of the Mean
• uncertainty in the average(sm);
different from the standard deviation
s (variation for each measurement);
if n=1, sm= s
• i. If s is known, the uncertainty in the
mean is:
• ii. If s is unknown, use the t-score to
compensate for the uncertainty in s.
• t - from a table for % confidence
level and n-1 degrees of freedom.
(one degree of freedom is used to
calculate the mean.)
Chebychev and Empirical
• 's Rule The proportion of observations within k
standard deviations of the mean, where , is at
least , i.e., at least 75%, 89%, and 94% of the
data are within 2, 3, and 4 standard deviations of
the mean, respectively.
• Empirical Rule If data follow a bell-shaped curve,
then approximately 68%, 95%, and 99.7% of the
data are within 1, 2, and 3 standard deviations of
the mean, respectively.
• -scores are a means of answering the question
``how many standard deviations away from the
mean is this observation?''
P 1sided
P 2sided
Confidence Interval
• The range of uncertainty
in a value at a stated
percent confidence
• Percent confidence… that
the value is within the
stated range
 s is known
 s is unknown
Look up the appropriate
z or t values to use:
x +/- t*s/sqrt(N)
Inferential Statistics
• Comparing two
sample means
T-test (Student's t)
• Used to calculate the confidence intervals of
a measurement when the population
standard deviation s is not known
• Used to compare two averages
• corrects for the uncertainty of the sample
standard deviation (s) caused by taking a
small number of samples.
Comparison Tests
Comparing the sample to the true value.
Comparing two experimental averages
Significance Testing
• Confidence interval
• Statistical Hypotheses
• Ho and H1
Comparison Test: Comparing the
sample to the true value
Method #1.
• If the difference between the measured value
and the true value (μ) is greater than the
uncertainty in the measurement, then there is a
significant difference between the two values at
that confidence level.
Method #2.
• experimental t-score (t ) is compared to t-critical
(found in a table)
• There is a significant difference if experimental
t is greater than critical t .
• t is chosen for N-1 degrees of freedom at the
desired percent confidence interval.
• If the experimental value may be greater or less
than the true value, use a two sided t-score. If …
Comparison Tests: Comparing two
experimental averages.
• use the pooled standard deviation and calculate t as:
• If experimental t is greater than critical t then there is
a significant difference between the two means.
• t is determined at the appropriate confidence level
from a table
• the t-statistic for N + N - 2 degrees of freedom.