What is statistical inference, Advanced Statistics

What is statistical inference?

 Statistical inference can be defined as the  method of drawing conclusions from data which are subject to random variations. This is based on the mathematical laws of probability. Probability is the branch of statistics, where we make inferences from a finite set of observations to an infinite set of new observations. The finite set of observations is called as Samples and the infinite set is called as populations. Suppose let us consider tossing a few coins. Here the total number of outcomes i.e., getting the heads and tails are called as the sample space. The new observation that is getting particularly heads or tails is the population.

Solutions to statistical inference:

There are two kinds of statistical inferences. One is the estimation, where we use the sample and sample variables to predict the population variables. The second is the hypothesis testing. Here we use the samples and sample variables to test the population and the population variables.


 Estimation can be divided into two types. One is point estimation and the other is interval estimation.  In the point estimation we use the sample variable to estimate the parameter ?, whereas in the interval estimation we use the samples variable to construct an interval which is equated to p where p is the confidence level adopted.

The most common method adopted for point estimation is the maximum likelihood estimation (MLE) which consists of choosing the estimate that maximizes the probability of the statistical material. MLE is the best solution if the statistical material is large. Special cases of MLE are the sample mean dented as E(X) and the relative frequency denoted by P(X=x).

Illustrations  of MLE:

A game is played with a single fair die. A player wins Rs.20 if a 2 turns up and Rs.40 if a 4 turns up, and he losses if a 6 turns up. While he neither wins nor loses if any other face turns up. Find the expected sum of money he can win.

Let X be the random variable denoting the amount he can win. The possible values are 20,40,-30,0

P[X=20] = P(getting 2) = 1/6

P[x=40] = P( getting 4 = )1/6

P[X = -30] = P(getting 6] = 1/6

The remaining probability is ½.

Hence the mean is E(X) = 20(1/6) + 40(1/6) + (-30)(1/6) +0(1/2) = 5.

Hence the expected sum of money he can win is Rs.5

Hypothetical Testing:

A statistical hypothesis is a statement of the numerical value of the population parameter.

The steps involved in solving a statistical hypothesis is

1.       State the null hypothesis Ho

2.       State the alternative hypothesis Ha

3.       Specify the level  of significance α

4.       Determine the critical regions and the appropriate test statistic.

5.       Compute the equivalent test statistic of the observed value of the parameter.

6.       Take the decision either to reject Ho or accept Ho.

Posted Date: 7/21/2012 9:16:51 AM | Location : United States

Related Discussions:- What is statistical inference, Assignment Help, Ask Question on What is statistical inference, Get Answer, Expert's Help, What is statistical inference Discussions

Write discussion on What is statistical inference
Your posts are moderated
Related Questions
explain the graphical method of measure of central tendency

We are installing a router for our network. We believe that the time between the arrival of packets will be exponentially distributed with parameter R = 2 packets/second, and th

Mauchly test is a test which a variance-covariance matrix of pair wise differences of responses in the set of longitudinal data is the scalar multiple of identity matrix, a proper

Paired availability design  is a design which can lessen selection bias in the situations where it is not possible to use random allocation of the subjects to treatments. The desig

Minimum volume ellipsoid is a term for ellipsoid of the minimum volume which covers some specified proportion of the set of multivariate data. It is commonly used to construct rob

Compliance : The extent to which the participants in a clinical trial follow trial protocol, for instance, following both the intervention regimen and trial procedures (clinical vi

Observation-driven model  is a term generally applied to models for the longitudinal data or time series which introduce within the unit correlation by specifying the conditional

Quasi-experiment is a term taken in use for studies which resemble experiments but are weak on some of the characteristics, particularly that allocation of the subjects to groups

It is an informal method of assessing the effect of the publication bias, generally in the context of the meta-analysis. The effect measures from each of the reported study are plo

Population averaged models are the models for kind of clustered data in which the marginal expectation of response variable is the main focus of interest. An alternative approach