You are on page 1of 3

1

2013, Study Session # 3, Reading # 10


SE = Standard Error
 = Rises SAMPLING & ESTIMATION  = Approaches to
RV = Random Variable
df = Degrees Of Freedom
CI = Class Interval
n = Sample Size

Sample Methods of Sampling Sampling error


A subgroup of Sample Corresponding
population. Statistic Population
Simple Random Stratified Random Systematic Parameter.
Sample Statistic Sampling Sampling Sampling
th
 It describes the  Each item of the  Uses a classification  Select every k
Sampling Distribution
characteristic of population under system. number.
Probability distribution of all
a sample. study has equal  Separates the  Resulting sample
possible sample statistics
 Sample statistic probability of population into strata should be
computed from a set of
itself is a random being selected. (small groups) based approximately
equal size samples
variable.  There is no on one or more random
randomly drawn.
guarantee of distinguishing
selection of characteristics.
items from a  Take random sample Standard Error (SE) of
particular from each stratum. Sample Mean
category.  It guarantees the  Standard deviation of
selection of items the distribution of
from a particular sample means.
category.
x =
n
 If is not known then;
s
Data Time sx =
Data Observational Characteristics n
Time series Observations Units  As n ; x approaches
taken over
equally spaced Longitudinal Same Multiple and
time interval S.E .
Panel Multiple Same
Cross- sectional Save time
Single point
estimate

Students T-Distribution
 Bell shaped.
 Shape is defined by df
 df is based on sample size.
 Symmetrical about its mean.
 Less peaked than normal distribution.
 Has fatter tails.
 More probability in tails i.e., more observations are
away from the centre of the distribution & more
outliers.

Copyright FinQuiz.com. All rights reserved.


2
2013, Study Session # 3, Reading # 10

Central Limit Theorem (CLT) Point Estimate (PE) Confidence Interval (CI)
For a random sample of size n with;  Single (sample) value Estimates
 population mean , used to estimate  Results in a range of values within
 finite variance (population population parameter. which actual parameter value will
variance divided by sample size)  fall.
 =
, the sampling distribution of
2
  PE (reliability factor SE).
sample mean x approaches a  = level of significance.
normal probability distribution Estimator: Formula used  1- = degree of confidence.
with mean & variance as n to compute PE.
becomes large.
Desirable properties of
Properties of CLT an estimator
 For n 30 sampling distribution
of mean is approx. normal. Unbiased Efficient Consistent
 Mean of distribution of all possible Expected value of If var ( ) < var ( ) As n , value of
samples = population mean . estimator equals of the same estimator
parameter e.g., parameter then  is approaches
 Variance of distribution =

E() = i.e, efficient parameter &
sampling error is than  sample error
zero. approaches 0
e.g., As n 
 &
CLT applies only when
SE  0
sample is random.

Copyright FinQuiz.com. All rights reserved.


3
2013, Study Session # 3, Reading # 10

Distribution Variance Sample Test Statistic Issues Regarding Selection


Non Small Large of Appropriate Sample Size
Normal Known Unknown t z
normal (n<30) (n30) As n ; s.e.  & hence C.I is
        narrow.
       
        Limitations of Large
      *  Sample Size
       
       
Large sample may Cost may
       
include increase more
      * 
observations from relative to an
more than one increase in
*The z-statistic is theoretically acceptable here, but use of the t-statistic is more population. precision.
conservative.

Biases

Data Mining Bias Sample Selection Bias Look ahead Bias Time-period Bias
Statistical significance of  Systematically excluding Using sample data that Time period over
the pattern is some data from analysis. wasnt available on the which the data is
overestimated because  It makes the sample test date. gathered is either
the results were found non-random. too short or too long.
through data mining.

Data Mining Survivorship Bias


Using the same data to  Most common form of
find patterns until the one sample selection bias.
that works is discovered.  Excluding weak
performances.
 Surviving sample is not
random.
Warning Signs of
Data Mining

Evidence of testing Lack of economic


many different, mostly theory consistent
unreported variables. with empirical
results.

Copyright FinQuiz.com. All rights reserved.

You might also like