Guide to Advanced Empirical


Analyzing Static Measurement Data



Download 1.5 Mb.
View original pdf
Page112/258
Date14.08.2024
Size1.5 Mb.
#64516
TypeGuide
1   ...   108   109   110   111   112   113   114   115   ...   258
2008-Guide to Advanced Empirical Software Engineering
3299771.3299772, BF01324126
4. Analyzing Static Measurement Data
4.1. Description
The first step in any statistical analysis is data description, and the first step of data description is to simply look at the data. Figure 2 shows the histograms for two different samples with the same mean and standard deviation without looking at these histograms, one would think from their descriptive statistics that both samples were from the same population. Looking at the distribution of values fora metric allows one to check for most frequent values (modes, outliers, and overall symmetry of the distribution. If a distribution is skewed by a few extreme values (large or small, many widely used statistics become misleading or invalid. For example, the mean and standard deviation are much more sensitive to extreme values than the median or percentiles, and so the mean of a skewed distribution will be far from the median and therefore a somewhat misleading measure of central tendency. Thus looking at the data allows us to determine which descriptive statistics are most appropriate.
As pointed out above, descriptive statistics such as point estimates are subject to error it is important to quantify this error so that the precision of the point estimate can be determined. The standard error of an estimate is a common way of representing


6 Statistical Methods and Measurement the precision of an estimate the range of values two standard errors on either side of the estimate delimit the 95% confidence interval for that estimate, i.e., the interval within which the true value of the parameter being estimated will fall 95% of the time. A wide confidence interval indicates that the estimate is not very precise, thus knowing the precision is useful for gauging an estimate’s value indecision making. The standard error increases as the sample size decreases, and the resulting imprecision in estimates is what makes very small samples so problematic.

Download 1.5 Mb.

Share with your friends:
1   ...   108   109   110   111   112   113   114   115   ...   258




The database is protected by copyright ©ininet.org 2024
send message

    Main page