Scribd
Upload
75 views

5 Errors

1. The document discusses sampling errors, non-sampling errors, standard error, and how to calculate sample sizes. 2. Standard error is the standard deviation of the sampling distribution of sample means and is used to calculate confidence intervals and for hypothesis testing. 3. The sample size needed depends on how close we want the sample estimate to be to the population parameter, the desired level of confidence, and the variability in the data. Larger sample sizes are needed for more heterogeneous populations.

Uploaded by

oliver sen
Copyright
© © All Rights Reserved
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
75 views

5 Errors

1. The document discusses sampling errors, non-sampling errors, standard error, and how to calculate sample sizes. 2. Standard error is the standard deviation of the sampling distribution of sample means and is used to calculate confidence intervals and for hypothesis testing. 3. The sample size needed depends on how close we want the sample estimate to be to the population parameter, the desired level of confidence, and the variability in the data. Larger sample sizes are needed for more heterogeneous populations.

Uploaded by

oliver sen
Copyright
© © All Rights Reserved
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
You are on page 1/ 21

Errors

Sampling errors
• Repeated samples from the same universe
• Different sample means
• This variation called sampling error
• Influenced by:
1. Sample size
2. Variability in reading values
Non sampling error
• Instrumental
• Observer
• Selection
Standard error
• Is the standard deviation of samples
distribution(many samples)
• It is derived from very large number of
samples, so it is more precise and smaller than
standard deviation
• It is used in hypothesis testing and calculation
of confidence interval
Standard error
• Drawing different samples give rise to
different sample means
• Frequency distribution of sample means give
rise to normal distribution of means around
the population mean (μ) or (p)

• SE of the means =

• SE of proportions=
Examples
SE of the means and CI
• Random sample of 25 males age 20-24 years
whose mean temp 98.14 F with SD = 0.6
• What we can say about the true mean of the
universe or general population?
• Use SE of mean
• 0.6/5= 0.12
• Put the confidence limits twice SE of the mean
based on normal distribution curve
• 98.14+ 2 SE x 0.12
• 2x0.12=0.24
• 98.14-0.24= 97.90
• 98.14+0.24= 98.38
• Confidence interval is 97.90 – 98.38 F
• From standard normal curve, we are confident
95% that the true population mean is within
these limits (two standard errors).
Example
SE of proportion
• 52% of population in a village are males
• Random sample of 100 give rise to 40% males
• Is this sample drawn from this population?
1. Use SE of proportion
2. Sq root of 52 x 48 /100 = 5
3. At two SE of p (95 confidence ) 52 + 2x5
4. 95% CI is : 42-62
5. So the sample p=40 is out side the limit
6. The difference not due to chance (significant difference)
7. The standard deviate = p-p^/SE= 40-52/5 =-2.4
Calculating CI
• When sample size of sample means large
• 95% CI for one sample mean of large samples
for scale data

• 95% CI for difference between two samples


means (large sample scale data)
• 95% CI for single proportion(p) in a large
sample
Sample Size
• In general, the larger the sample size (selected
with the use of probability techniques) the
better.
• The more heterogeneous a population is on a
variety of characteristics (e.g. race, age, sexual
orientation, religion) then a larger sample is
needed to reflect that diversity.
• Response rates vary on the type of surveys
(e.g. mail surveys, telephone surveys).

12
Binomial distribution Normal distribution
P µ Center of normal curve
Mean of red balls
 Standard deviation
of normal curve
= /
Standard error of the Standard error of the
proportions means

Observed sample outcome

Standard normal deviate


(z) Of one sample mean

z for sampling distribution


=Z of sample means
For binomial data
• Z=
• Z=
• where is the observed sample proportion and
we wish to solve for n
1. How close the will be from p = -p= δ
0.04, 0.05!!
2. How sure the will be close within δ of the true
p.
suppose 95% certain that will be at δ distance of
true p ( z at 0.05 = 2~1.96)
3. What the anticipation of p value? If not found
put as 0.5
• in the equation
z=
Example
• Suppose that you want to estimate the
knowledge of mothers about malnutrition in
community X with 95% confidence. Assume
the proportion from a previous study in near
country is 0.5 Kg. Find the required sample
size.
• P= 0.5
• q = 1-p= 1.0.5 = 0.5
• Z at 95% CI =1.96
• = error = (0.05)

• n= ???
For quantitative data
• Z=
• How close is the sample mean from
population mean = δ( marginal
error)
• How sure the sample mean will be close
within δ of the true mean (μ) at z=1.96~2

•
Example:
Suppose that you want to estimate average birth-weight
of children born to non-smoker women in community X
with 95% confidence. Assume the standard deviation of
birth-weights is 0.5 Kg. Find the required sample size.

[(1.96)( 0.5)] 2
n 2
 96
(0.01)
Therefore in this case a sample size of at least 96 is required

You might also like