Scribd
Upload
3 views

Estimation

The document discusses the concept of estimation in inferential statistics, emphasizing that sample means can vary from the population mean and that estimation is used to predict unknown population parameters. It explains the importance of confidence intervals, which provide a range within which the true population mean is likely to fall, and highlights the trade-off between the width of the interval and the confidence level. An example is provided to illustrate how to construct a confidence interval for a sample mean, demonstrating the calculation process and the interpretation of the results.

Uploaded by

samueljuma2004
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
3 views

Estimation

The document discusses the concept of estimation in inferential statistics, emphasizing that sample means can vary from the population mean and that estimation is used to predict unknown population parameters. It explains the importance of confidence intervals, which provide a range within which the true population mean is likely to fall, and highlights the trade-off between the width of the interval and the confidence level. An example is provided to illustrate how to construct a confidence interval for a sample mean, demonstrating the calculation process and the interpretation of the results.

Uploaded by

samueljuma2004
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 14

Estimation

INTRODUCTION
From sampling distributions we noted that not all sample means
(x̅) are equal to the population mean (𝜇)
In the example of 10 objects weighing 1—10 Kgs we drew ten
systematic samples and calculated means of every sample
• We noted that sample means vary from
the lowest of 2.5 Kg to the highest of 8.5
Kg
• Only serials 4 and 9 were equal to the
population mean
• Therefore it means that for any one
sample mean to be equal to the
population mean it becomes a matter of
probability
• Our example is a bit simplistic because the population size (10
objects) is too small that parameter, i.e. the mean (𝜇) can be easily
computed by adding up the weights and dividing by the known size.
• However in real life situations the population size is so large that it
would take longer and tedious work to enumerate all the members in
order to compute a parameter like the mean.
• Therefore in inferential statistics we use sample statistics (x̅, S) to
predict the unknown population parameters e.g. (𝜇, σ)
• Estimation is a statistical procedure for predicting the value of the
mean (𝜇) of a population based on a known statistical attribute (x̅) of
one sample
• Whereas our previous example had n = 4 and N = 10, in inferential
statistics sample size must be > 30 and population must be larger
• In estimation we take into account the fact that for a randomly selected
sample based on adequate sample size, the sample mean will be close to
the population mean
• The question in Estimation therefore is how close?
• Example: assuming in this class of inferential statistics, a sample of 36
students is taken an average weight of 56 Kg is computed with a standard
deviation of 3 Kg. let us use the provided statistics to predict the
population mean
• We can assume that the best point estimate of 𝜇 from the sample statistics
is 56, since the sample is randomly drawn and the size n > 30.
• However it is unrealistic to assume that the mean of a large population will
be exactly equal to the mean of a sample drawn from it
• Since a point estimate would not suffice, we can do an interval estimate
and conclude that the mean lies within a chosen interval
• For instance we can guess that the approximate estimate of 𝜇 lies in
the interval of
i. 55.5—56.5
ii. or 55—57
iii. or 54—58
iv. or 50—62.
• In each case the interval is centered at x̅ = 56
• The first possibility is more accurate prediction than the point
estimate of 56 Kgs but it is a bit too restricted that it may not be
possible that the true mean 𝜇 will be that close to x̅ = 56
• So if this were the prediction you would not have a high degree of
confidence in its accuracy
• The second interval 55—57 is a better prediction in that it is more
likely it will contain the population 𝜇 hence we would have a greater
confidence in it
• Note that as the length of the estimating interval increases, our
degree of confidence in its actually containing the population mean 𝜇
would also be increasing
• At the extreme 50—62 we become certain that 𝜇 lies within the
interval because it is such a wide interval that there would be little
doubt that 𝜇 is within it
• In this case we could have reached this conclusion without the
bother of collecting data from a sample
• The interval is so large that using it as a prediction of the 𝜇 is
worthless
• Therefore if an interval is too large, then the corresponding
prediction is worthless
• To obtain a useful estimate we must avoid the extremes: 55.5—56.5
too narrow and 𝜇 is likely to miss out while 50—72 too wide and
meaningless
• So we must settle to an intermediate interval but when we do so we
must accept that there is a chance of being wrong
• How large a chance of error are we willing to accept, 90%, 95%, 99%
of being right?
• The percentage is known as the confidence level of the estimate
• The corresponding interval is called confidence interval
• Suppose we select a confidence level of 95%.
• That is to say we want to construct a confidence interval centered at
the sample mean x̅ which contains the true population mean 𝜇 with
95% certainty.
• In other words, this means that the probability is .95 that the 95%
confidence interval actually contains the unknown mean 𝜇.
• The figure of 95% also reminds us of one of the primary
characteristics of the normal distribution i.e. 95% of the population
fall within 2 standard deviation
Most common confidence levels
Confidence Level (%) Z value
90 1.64 (1.65)
95 1.96
98 2.33
99 2.58
• The confidence interval for the mean 𝜇 is an interval centred at the
sample mean x̅ and extending z.σx̅ in both directions. The confidence
interval is given by
• x̅ ± zσx̅ = [x̅ - zσx̅, x̅ + zσx̅ ]
• Example: Construct a 95% confidence interval for the mean 𝜇 height
of students based on the sample of size n = 36 with sample mean x̅ =
66 inches. Assume the standard deviation for adult heights is 3
inches.
• Solution: we are given
• n = 36,
• x̅ = 66,
• σ = 3,
• Since sample size is greater than 30, 𝜇 = 𝜇x̅
σ 3 3
• And Standard Error: σx̅ = = = = 0.5
𝑛 36 6
• 95% has a z value of 1.96
• Use the formula: x̅ ± zσx̅ = [x̅ - zσx̅, x̅ + zσx̅ ]
• 66 ± 1.96(.5)
• 66 ± .98
• [65.02, 66.98]

• At 95% confidence or certainty we conclude that the true population


mean 𝜇 is between 65.02 and 66.98 inches.
• Equivalently the probability that this interval contains 𝜇 is .95.
• Exercise: Construct a 99% confidence interval for the mean height of
students based on the same data on students heights.

You might also like