Nonparametric Tests

Yates correction calculation

((Observed-Expected) - .5)2 / Expected

How is nominal data evaluated?

-Can't compare mean, median or mode -Can only look at frequencies

How is data ranked in nonparametric tests?

-Data is arranged from smallest to largest -Smallest value given a rank of 1

When are distributions nonparametric?

-Pathologic conditions -Small clinical samples -Samples of convenience

What is the difference between skewed and rectangular distributions?

-Tails are differently shaped -Not equal area under opposites of the median

T/F The Mann-Whitney U test requires that the groups be the same size

False - does not

T/F The Wilcoxon Signed Ranks Test requires that the groups be the same size

False - does not

T/F The Fischer Exact test does not give a X2 value an a p value

False - it gives both

T/F Nonparametric tests have higher statistical power than parametric tests

False - lower

T/F With nonparametric tests you can use a smaller sample size to show significance than with parametric tests

False - may need larger sample size

T/F To use nonparametric tests you must have at least ordinal, interval, ratio or nominal data

False - not appropriate for nominal data

T/F Chi squared frequencies represent ranks or percentages

False - only represent individual counts

Alternative to the Yates correction for those who think it is too conservative

Fischer Exact Test

Which test is the Mann-Whitney U test analogous to?

Independent t-test

Nonparametric test used with independent groups

Mann-Whitney U Test

Nonparametric Tests

Mann-Whitney U Test Wilcoxon Signed Ranks Test

Version of X2 used with paired data

McNemar Test

Types of nonparametric distributions

Skewed Rectangular

In the Wilcoxon Signed Ranks Test the ___ is calculated, which measures the number of _____ above and below the median

T statistic, paired ranks

How is data ranked in nonparametric tests when there are tied values?

Tied values are given equal ranks which are an average of the places they would fill (eg. if the two values are equal and would fill the 4th and 5th rank, they are both given the rank of 4.5, and the next value would be given a rank of 6)

T/F Nonparametric tests use ranking of data rather than using precise measurements

True

Nonparametric test used with paired groups

Wilcoxon Signed Ranks Test

When doing X2 on a computer will calculate a p value for the calculated __ rather than have you compare the X2 to the ___

X2, critical value

Yates correction decreases the value of __, making it more difficult to show a ____

X2, statistical difference

What happens to the chi square value as the observed becomes more different than the expected?

chi squared rises

Test used to analyze frequencies

chi squared test

Expected frequencies can be from a ___ sample, or calculated by assigning ___ probabilities to each category

control, equal

The Fisher Exact Test is a more complicated test that calculates _____ given the observed frequencies

exact probabilities

Characteristic of chi squared categories

exhaustive and mutually exclusive (data not assigned to more than one category)

In the Mann-Whitney U test a (one/two)-tailed p value is given, and compared with _____

two, pre-selected critical level for p (e.g. p = .05)

The Wilcoxon Signed Ranks Test assumes variance is ____

unequal

Chi Squared is less reliable with ___ sample sizes

very small

U is converted to a standard ____ for each group

z-score

In the Wilcoxon Signed Ranks Test the T is converted to a standard ____ and a ___ is calculated and compared to the ____

z-score, p value, pre-selected critical value of p

Chi square value when observed = expected frequency

zero

In a Yates correction you subtract __ from each O-E calculation

0.5

With chi squared no more than 20% of the cells should contain expected frequencies of less than __

5

What do you do if more than 20% of the expected cell frequencies are less than 5?

Apply the Yates Correction

How to increase the expected frequency with small sample sizes

Combine categories

When doing X2 on a computer you must put in ____ values

individual

U is the measure of how often ranks in one group are ____ than ranks in the other group

lower

Mann-Whitney U Test is used to compare the ___ of two independent samples

medians

Wilcoxon Signed Ranks Test compares the ____ of two matched groups

medians

The chi squared tests assumes all data is _____

nonparametric

What information is needed to use the chi squared test?

observed frequency and the expected frequency

When doing X2 on a computer you will get a ____ for the entire table and ___ for each category

p value, X2

Which test is the Wilcoxon Signed Ranks Test analagous to?

paired t-test

When doing X2 on a computer you can compare X2 to another ____ or to the ___ of the entire data group

sample (control), mean

In the Wilcoxon Signed Ranks Test the data is put in order and given ____ ranks comparing them to the group median

signed (+/-)