Ch 2: Describing Location in a Distribution
Part 1: Measures of Relative Standing and Density Curves
- Shows relatives standings with numbers
- if distribution is roughly symmetric with no obvious outliers... mean is around the same as the average
- we can calculate with z-scores
- x-score = (x-mean)/(standard deviation)
- called standardizing, so the z-score is also called the standardized value
- tells how many standard deviations away from mean a value is
- can be used to compare numbers
- z-scores to percentiles
- can use Chebyshev’s inequality
- can be used in any distribution
- percent of observations falling within k standard deviations from the mean is at least (100)(1-1/k2)
- for example, if k = 1, at least 0% of observations will fall within 1 standard deviation of the mean
- shows that it is unusual to find observation more than 5 standard deviations from the mean
- gives us insight into how observations are distributed, but not percentiles; need density curves to do that
- density curves: always on or above horizontal axis and has area of 1
- the curve better to use than histograms because does not depend on our choices
- describes overall pattern of distribution; doesn’t accurately describe data, but accurately enough for use
- come in many shapes; one kind is a normal shape
- Mean and median in density curves
- median: equal-areas point, and if the curve is Normal, then median is its center
- mean: balancing point of any(?) distribution, including density curves
- in a symmetric curve, these two are equal; in a Normal curve, then these two are approximately equal (?)
- notations
- X-bar and s for mean and stan. dev. in actual observations (data), M and O-- for density curves
Part 2: Normal Distributions
- Normal distributions: symmetrical, bell-shaped density curves; they are special curves
- all have similar shape, and are described by M (mean) and O-- (standard deviation)
- N(Mean, Stan. dev.)
- 68-95-99.7 rule: around 68% observations fall within 1 stan. dev. from mean, around 95% observations fall within 2 stan. dev. from mean, and around 99.7% observations fall within 3 stan. dev. from mean
- more precise than Chebyshev’s rule
- standard Normal distribution: distribution of the z-score, if the original variable also have a normal distribution
- standardizing is also a linear transformation
- can calculate percentiles using Table A, or the standard Normal table
- calculate z-score to two decimal places, then use the table to find the percentage to the left of the z-score (less than the z-score or the indicated value)
- if want to calculate percentage more than the value, then use 1 - indicated percentage
- to answer question, answer in context and is reasonable
- ex: If Sasha got a score of x in the test that has the normal distribution of N(M, O--), what is her percentile? Answer: Sasha’s percentile will be ___ if she got a score of x.
- always sketch the curve, mark z-score, then shade area of interest
- Summary of calculating
- find what the question is asking and what variable it is stating and draw and shade distribution; calculate z-score and mark that
- use the table and find corresponding percents
- answer question in context
- if the z-scores exceed the values given by Table A, then take the Table’s percentages as 0 (if the scores are negative) or 100 (if the scores are positive)
- Finding value given the proportion
- the process is basically just the backwards of finding the proportion
- find corresponding proportion, then find the z-score, then use the z-score equation to solve for x
- Determining if the data is Normal
- make histogram or stemplot and see if graph is around bell-shaped and symmetric around mean
- use normality plot; if lie in straight line, then data is Normal; if a point is far away from overall pattern, then it is an outlier
- to do this, plot the z-scores on the x-axis, and data on the y-axis...
- turn on a plot, and select the lower-last picture of a plot; that is the normality plot
- if the plot is not linear...
- if the largest observations are distinctly above the line drawn through the main body of plots, then it is skewed right
- if smallest observations distinctly below the line drawn through main body of plots, then it is skewed left
- only concentrate on shapes that are distinctly not linear, and ignore the wiggles
more vocab and concepts
No comments:
Post a Comment