Please enable JavaScript.

Coggle requires JavaScript to display documents.

ML (Learn about data (Dimensions & attributes (nb of variable : ncol,…

- - - - test for Missing values
      - recode
      - Exclude
  - - - numerical data
        
        central tendency
        
        Mean
        
        Mode
        
        Median
        
        calcul for indiv variable/ across all variables in a particular data frame
        
        variability
        
        range
        
        percentiles
        
        fivenum() provides min, 25%, 50%, 75%,max
        
        quantile() 0%, 25%, 50%, 75%; 100%
        
        custom :red_flag: quantile(mtcars$mpg, probs = seq(from = 0, to = 1, by = .1))
        
        Variances
        
        shapes
        
        skewness: tail (right/left)
        
        kurtosis : flat/ nhon
        
        Outliers
        
        Distort predictions, affect accuracy
        if outliers present, which obsers are outliers
        
        outlier() gets the most extreme observ from mean
        scores() computes the normalized (z,t,chisq, etc...) score to find observ lie beyond a given value:
        observations that are outliers based on z-scores greater than 2. In other words, these observations exceed 2 standard deviations from the mean.
        z_scores <- scores(mtcars$mpg, type = "z")
        which(abs(z_scores) > 2)
        
        outliers based on values less than or greater than the "whiskers" on a
        boxplot (1.5 x IQR or more below 1st quartile or above 3rd quartile)
        which(scores(mtcars$mpg, type = "iqr", lim = 1.5))