Non-normal distributions

Question 1

Name 2 methods of identifying outliers?

Answer 1

1. Visual inspection of histograms and box plots

2. SPSS - convert scores to Z scores

Question 2

How do we treat outliers?

Answer 2

1. delete the outlier ( not advised on small samples as may reduce statistical power)
2. change outlier score to one large than the next highest score ( retains sample size, rank maintained, reduces tails of distribution to improve normality)

Question 3

How to identify non-normality?

Answer 3

1. Visual inspection of a histogram, Q Q plots ( squiggle line), stem and leaf plots, box plot ( check whiskers)
2. SPSS tests - Kolmogrov Smirnov and Shapiro wilk ( sig if less than .05 meaning problem with normality)
3. SPSS descriptives output - Z skew and z kurtosis ( and no. away from 0 could be problem + pos skew - neg skew. have to hand calculate into z scores - statistic /error = if outside +-3.33 their is a problem with normality alpha p

Question 4

How to treat non - normality?

Answer 4

1. Use Non -parametric tests ( chi-square, Kruskal Wallis, Friedmans)
2. Transformation of raw scores in SPSS ( long process to improve distribution
3. Bootstrapping - Resampling with replacement 5000 times in SPSS . Creates normal-shaped distribution.

Question 5

Median, mode and the mean - where do they lie in the skew?

Answer 5

Median always in the middle, mean is at the tail and mode is the highest score

Question 6

Skeweness - High (positive) , normal and low ( negative) Name them?

Answer 6

Leptokurtosis = high - positive
Mesokurtosis = normal
Platykurtosis =  low - negative