Medical statistics Key Areas

Question 1

Primary vs Secondary data

Answer 1

Primary Data (Indicative):

• Indication/ impression of reality but difficiult to analyse
• This is Blood. Tissue samples, patient records, histological slide, DNA sequence, fingerprint, cadaver

Secondary Data (Derivative):

• Numbers derived from primary data and easier to analyse
• Peak response, BP/ counts, surface coverage, antibody levels, weight/age, number of deaths

Question 2

Normal distribution

Answer 2

Question 3

Why is controllign osurces of variability important in studies

Answer 3

‘Overall’ mean and SD substantially misrepresent the real world. Hence contorlling sources of varibaility is key to good studies so finding out values for both males and females etc.

Question 4

What is Experimential design about?

Answer 4

• Identifying
• Defining
• Mnaaging
• Quantifying

Source sof variability can be what you introduce, plain random etc or groups.

Question 5

WHat do parameters and error bars tell us?

Answer 5

Question 6

Descriptive and inferential measures of a sample

Answer 6

Question 7

WHta is Simpsons Paradox?

Answer 7

• Analysis must match design
• Design must match hypothesis
• Be suspicious of dichotomisstion (small/large)

Question 8

Principles of experimential design

Answer 8

• Controlling eg, placebo
• Rnadomising - systemic bias
• blinding -personal bias
• Replication - increase precision (Sample size and statistical power)

Question 9

Important questions vs Trivial Questions in Hypothesis testing

Answer 9

Important questions - easy to ask, difficult to answer

Trivisal questions - easy to ask, easy to answer

Question 10

Null VS Alternative hypothesis

Answer 10

Null hypothesis = population parameter is equal to hypothesised value.

. Alterntive hypothesis = You might believe to be truw or hope to prove true

Question 11

Family- Wise Error Rate (FWER)

Answer 11

Question 12

Type II errors

Answer 12

Question 13

P- value

Answer 13

Informally = Probability under specified statisticla model that a statistical summary of the date (eg sample mena difference between 2 comapred groups) would be equal to or more etreme than its observed value.

• ALWAYS refers to NULL hypothesis

Question 14

WHat is. t TEST?

Answer 14

When Null hypothesis is true on average t=0

Question 15

T test and graph

Answer 15

On avergae t is not 0 when null hypothesis is NOT true

Question 16

Interpreting P values

Answer 16

Question 17

Power Analysis (6 related variables)

Answer 17

Question 18

1 vs 2 tailed test

Answer 18

Question 19

COnfidence intervals VS Credible itnerval

Answer 19

A confidence interval gives an indication of the degree of uncertainty of an estimate, showing the precision of a sample estimate. The 95% confience intervals are calculated so that if we repeated the study many times, 95% of the time the true unknown value would lie between lower and upper ocnfidence limits. A wider interval indicates more uncertainty in the estimate. Overlapping confidence intervals indicate that there may not be true difference between 2 estimates.

Credible interval = gives indication of ucnertainty of an estimate from data anlysis. 95% credible intervals are calculated so that there is 95% probability of the true value lying in the interval.

Question 20

P vales, CI and standard errors of means (differences etc)

Answer 20

Question 21

Sampling Distirbution

Answer 21

• Probability distribution of a sample statistic eg, mean
• The standard deviation of the sampling sistirbution is the standard error of the sample statistic for sample size = n
• As n increases, the sampling distribution (and therefore its standard deviation) narrows
• By a miracle of statistics, if is often possible to estimate the SE from a single sample with n data…

Question 22

H0, sem and CIs

Answer 22

Question 23

Normal vs t distribtuons

Answer 23

Question 24

H0, sem and CIs example 2

Answer 24

Question 25

P values vs confidence intervals

Answer 25

If p = 0.5 then one of the 95% CIs will = 0

If p= a then one of the CIs will equal uo

Question 26

Forets plot example

Answer 26

The horizontal line is the CI so wider then wider CI and size dot is usually proportional to size sample so you can usually see the large sample, narrower confidence interval.

The diamond is the meta analysis