Basic Statistics

Question 1

Define the term accuracy

Answer 1

How well a measurement matches exactly an accepted standard

Question 2

Define precision

Answer 2

When there is a small random error of estimation

Question 3

Define reliability

Answer 3

One that can be repeated with minimal variation

Question 4

Define validity

Answer 4

When something gives genuine information about that which is being measured.

Question 5

List 2 tests used to test for data set normality

Answer 5

1) Shapiro-Wilk test

2) Kolmogorov-Smirnov test

Question 6

When do you use mean for datasets, and when do you use median?

Answer 6

Normally distributed data = Mean

Non-normally distributed data = Median

Question 7

Define variance

Answer 7

Variance = The average amount by which any individual measurement differs from the mean

Question 8

Define standard deviation, and how it is calculated

Answer 8

STDV = The square root of the variance.

Question 9

Define what the coefficient of variation is

Answer 9

Coefficient of variation = Describes the STDV as a percentage of the mean

COV = stdv/mean x 100

Question 10

Define the ‘standard error of the mean’

Answer 10

Standard error of the mean = Gives information on how close a sample mean is to the actual population mean.

SEM = S / Square root of (n)

Question 11

Define a confidence interval

Answer 11

This describes the distribution of data in the sample
This can only be used in normally distributed samples with more the n=30

Example :
Say mean is 52, and SEM is 3kg.

95% CI = 52 +/- (2x3)
OR
= 52 +/- 6
OR
=46-58

So you calculate the confidence interval using the mean and the SEM.

Question 12

How can you transform non-normal or non-parametric data?

Answer 12

Can be transformed into normally distributed data.

This can be done in several ways:

1) Taking the logarithm out of the data
2) Taking the square root
3) Squaring
4) Taking the reciprocal (1/x)

Question 13

What is binomial distribution and when is it used?

Answer 13

Binomial distribution = Simplest form. Used for data measured on a discontinuous dischomaous scale.

Example = Yes/No, Dead/Alive

Question 14

What is poisson distribution and when is it used?

Answer 14

Poisson distribution = Used for discrete data, when number of occurences of an event per unit of time are counted.

Example = NUmber of admissions to gynae unit from ED each day.

Poisson = Allows you to look at the probability of admission every day, so is useful for when considering random sums of rare events.

Question 15

How do you calculate stdv in poisson distributions?

Answer 15

In poisson distribution

Stdv = Square root of the mean

Question 16

How do you calculate the SEM or standard error of the mean in binomial distibutions?

Answer 16

P = Patients in one category
Q = (1-p) = Or the patients in the other category

SEM = Square root of: ( P*Q / n )

Question 17

Describe type 1 and type 2 errors

Answer 17

Alpha or type 1 errors = When null hypothesis is wrongly rejected. So when P is over 0.05

Beta or type 2 errors = When null hypothesis is wrongly NOT rejected.

Question 18

Define power

Answer 18

Power of a study = Ability to detect an effect of a specified size.

Power = (1-b) OR 100 (1-B)%

Usually you calculate the N in order for the power to be between 80-90%.

Question 19

List 7 statistical tests that can be used in non-parametric data and briefly describe when they are used

Answer 19

1) Sign test
2) Chi Square of X2 test = For 2x2 and larger tables.
3) McNemar X2 test = For before and after categoricla data
4) = Wilcoxin test = For paired data. The same same subject in different conditions
5) Mann-Whitney U test = For unpaired data
6) = Kruskal-Wallis analysis for variance
7) Spearman’s Rho = For testing association between groups

Question 20

Name 6 statistical tests that can be used in parametric data, and briefly when they are used

Answer 20

1) Student’s T test = For paired and unpaired data.
2) F test = To determine equality of variances
3) Analysis of variance = Can be one way or 2 or more way. For repeated measures
4) Pearsons R = For associations between groups
5) Linear regression analysis = Also for associations
6) Multiple regression analysis = Also for associations.

Question 21

Define sensitivity

Answer 21

% of true positives

Question 22

Define specificity

Answer 22

% of true negatives

Question 23

Define positive predictive value

Answer 23

% of people with positive results who are correctly diagnosed

PPV = True positives/All cases

Question 24

Define negative predictive value

Answer 24

% of people with negative results that were correctly diagnosed.

NPV = True negatives/All cases

Question 25

Define the odds ratio and how it is calculated

Answer 25

Odds ratio = Gives an estimate for the association between two binary (yes/no) variables.

• It allows you to test the effects of other variables on that association (logistic regression)
• You can also calculate a confidence interval with the OR

Calculation = Number of people with outcome / Number of people without outcome

Or A / B

Question 26

Define relative risk and how it is clacualted

Answer 26

Relative risk = Compares the proportions of people with a particular outcome in two groups.
-Can also calculate the confidence interval with RR

So difference between RR and OR = Is that in RR you are dividing the proportions instead of raw numbers like in OR

RR = (A / (A+b)) / (C / (C+D))