features we always hope to explore the overall data from some sample data, so it is in the website data analysis, we try to show in recent days of data to predict the current overall situation of the web is like, there is no signal for better or worse, but the days of data can not be completely representative of the general population so it can only use "estimation". At the same time, the website data will always exist fluctuations in recent periods of data as the sample data is likely to be just in the lower or higher level, so we estimate sample value obtained is not possible without bias, we also need to assess the estimated change interval can be.
parameter estimation (Parameter, Estimation) refers to the method of estimating population parameters using sample statistics, including point estimation and interval estimation.
point estimation (Point, Estimation) is a method of statistical inference, which uses the sampled statistical index as the estimation of the eigenvalues of an unknown parameter of the whole.
general on the overall estimation of the parameters will include two categories: one is to use the sample mean to estimate the population mean, corresponding to the numeric index website data, such as web UV every day, we can use nearly a week of daily UV to estimate the general situation of the number of unique visitors to the site every day; the other one is with the sample probability to estimate the overall probability, corresponding to the ratio index website data, such as the goal of the website conversion rate, we can use the conversion rate of nearly 3 days to the estimated level of target transformation site that day and we will calculate the sample; the standard deviation to explain the fluctuation of sample mean and probability to size. Estimation of fluctuations in the overall data.
point estimation also includes the use of least squares method for linear regression to do curve parameter fitting, and maximum likelihood estimation method to calculate the probability density function of the sample set distribution parameters.
interval estimation (Interval, Estimation) is based on the extracted samples, according to a certain degree of accuracy and accuracy requirements, the overall estimate of the unknown parameters may be the value range. The interval estimation is generally calculate confidence intervals or get the overall average overall probability at a given confidence level (Confidence Interval), usually according to the sample number and the standard deviation of the overall standard error calculated, according to the point estimates for the sample mean and sample probability estimation of population mean or total probability, and then draw a a value of the upper and lower critical point.
, we can take the standard deviation of the sample as S. If we sample n samples, then the standard sigma of the population can be estimated by the standard deviation of the sample