applied statistics
Inferential statistics need to describe the overall sample, there should be statistics. Note that you must generally normal distribution, or distribution of statistics can not be obtained. Chi-square distribution and t distribution as long as the sample is greater than 30 are approximately normal.
t distribution and F distribution derivation and application (FIG):
The overall proportion of π, the proportion of the sample is p ratio can be used to calculate prevalence. Approximation is the mean and variance does not change, but the distribution in the form of change, but in fact did not change shape. EG : When large sample, the binomial distribution approximates the normal distribution:
Unbiased sample of the first moment of use. Effectiveness of Samples second moment, it can be seen better average than the median. Consistency use sample third moment, the statistics are in line with the general appearance.
Point estimate is computed directly does not require the distribution of the sample mean and variance; estimated interval is calculated as a distribution section
Relationship between the point estimation section estimates, as shown below:
Confidence interval: 100 times 95 within the time interval has true value.
Practical applications, the standard deviation can be replaced by historical values.
The mean standard deviation has the same units, the dimensions may be removed, as a numerical value, EG : P-value .
General expressed confidence interval (point estimate + , - the value of a distribution * statistic standard error), we can see from the expression you want to increase the length of the confidence interval confidence, but does not increase the interval, need to increase the sample size. One end of one side only care about the value interval available.