12 June 2021 18:45 Data Analysis There are 3 types of data analysis. GAUSSIAN: Analyse Gaussian (normally-distributed) data. The response variable is usually continuous data which isn't bounded. POISSON: Analyse Poisson data. The response variable is usually integer (count) data which is bounded below at zero. BINOMIAL: Analyse binomial data. The response variable is usually specified as two integer columns representing 'successes' (an event occurring), and 'failures' (an event not occurring), where the total number of 'trials' (successes + failures) can be different on each row. T Data visualisation is a set of different layers of information. · Data is the most fundamental base layer. All figures are plotted from data. · Aesthetics describe the mapping of data onto the graph. Foremost, aesthetics are used to state what the x-variable and y-variable are, although they can extend to other variables for more complex graphs (e.g. a third variable to colour data points, or a third axis for a 3D graph). · Geometries describe how aesthetics are represented. This principally includes the graph "type" (e.g. histogram, boxplot, barplot, scatter plot etc.). · Facets can be used to split a figure into sub-plots. Sometimes putting all data in one space is difficult to read, so a plot can be split out into different facets according to chosen aesthetics. · Statistics describe any models used to describe or predict the data. This may commonly include means and standard deviations, linear models or loess curves. · Coordinates are not commonly used for simple graphs, but are useful when projecting the data into a different coordinate space. Mapping is the most obvious example: the same data can be plotted using many different geographic coordinate systems. Three most important: 1) Data 2) Aesthetic 3) Geometry DATA DISTRIBUTIONS The different types of data: . numeric (quantitative) o continuous: Gaussian, beta, exponential o discrete (counts): Poisson, binomial . non-numeric (qualitative) o categorical o ordinal Distributions also show us how variables are: The distributions of variables are important factors in how we interpret plots and analyses. Distributions are effectively 'counts' of the number of times certain numeric variable values occur. - For discrete numeric variables, such as the number of countries or continents, the distribution may be simply the counts of occurrence of 0, 1, 2, 3, 4, etc. - For both discrete and continuous numeric variables, the distribution represents the number of times variable values occur within certain intervals (e.g. for the eyesight variable we can count the number of values between -3 to -2, -2 to -1, -1 to 0, ... , or any other intervals we choose). · bounded (i.e. cannot occur above, below or between certain values) · how they are skewed (e.g. there are lots of small values but not many large ones or vice versa) · what values their moments (i.e. mean, variance, skewness) may take · which statistical distributions the data approximate. Distributions of your data are important, as all of the things listed here can have implications for: · how you