Brainscape's Knowledge GenomeTM

Browse over 1 million classes created by top students, professors, publishers, and experts.


See full index

MFSRH Part 2 2024 > STATS > Flashcards

STATS Flashcards

1
Q

What kind of data is the mean helpful in interpreting?

A

Normally distributed data. If there is an outlier the mean is less helpful as it skews the mean so it is no longer typical of the majority of the data

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
2
Q

What kind of data is the median helpful in representing?

A

When data are not normally distributed ie they are skewed.
Half the data will be above the median and half will be below.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
3
Q

In relation to the median what does a box and whisker plot show?

A

It gives the interquartile range (mark point where a quarter of the data lows below and where 3/4 data lies below in a box-whiskers extend out to the limits of the data) it shows where half of the patient data lies around the median

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
4
Q

What information does the standard deviation provide you with?

A

It tells you in a normally distributed data set, how much the data is clustered or spread out around the mean

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
5
Q

How much of the normally distributed data lies within 1 SD above or below the mean?

A

68.2%

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
6
Q

How much of the normally distributed data lies within 2 SD above or below the mean?

A

95.4%

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
7
Q

How much of the normally distributed data lies within 3 SD above or below the mean?

A

99.7%

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
8
Q

How can you use the SD to check if data is normally distributed?

A

look at the dataset. if you apply 2 SD to the mean if it is normally distributed you will not have impossible values/values that were not detected in it

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
9
Q

What does a level of agreement >0.5 indicate between data sets?

A

That there is agreement

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
10
Q

Which p values are significant?

A

P =0.01 and P =0.001
Likelihood result due to chance is 1:100/1:1000

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
11
Q

What does sensitivity measure?

A

How good a screening test is at identifying disease when present ‘the pick up’

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
12
Q

What does Specificity measure?

A

How well a screening test identifies healthy people correctly ‘negative in health’

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
13
Q

How do you work out sensitivity?

A

True positive / everyone who has the disease

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
14
Q

How do you work out specificity?

A

True negative/everyone who doesn’t have disease

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
15
Q

What is positive predictive value?

A

If result positive how likely it patient has the disease?

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
16
Q

How do you calculate PPV?

A

True positive/everyone who tested positive

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
17
Q

What is negative predictive value?

A

Likelihood that if test negative it’s a true negative

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
18
Q

How do you calculate NPV?

A

True negative/all the negatives

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
19
Q

What is the likelihood ratio?

A

Tells you have much performing a test adds to the pre test probability of pt having (or not having) a condition

Positive likelihood ratio =Higher the number the more likely a positive test indicates the condition.

Closer to 1-test adds little to your pre test knowledge of probability of disease

The Negative likelihood ratio means how much less likely it is pt has disease compare to pre test probability

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
20
Q

What are risk, odds, nnt trying to help you with?

A

Understanding which sample did better or worse. Making comparisons.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
21
Q

What are confidence intervals trying to help you with?

A

Calculating what the true target population results are without having to experiment on whole target population

In order to be accurate selection bias must be minimal

The 95% confidence interval for a population mean, originates from the standard deviation eg it describes 2 SD (95.4% data) around the observed mean ie the range you are 95% sure the population result lies

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
22
Q

What are the mean/mode/median useful for?

A

Describing the sample population results rather than listing each individual outcome

To be accurate must be minimal observation bias

23
Q

How does a p value help?

A

You described your results

You estimated what the true results would be if had sampled whole population (CI)

You’ve analysed your sample results and worked out your risk ratio etc for your samples

Now you use statistics to analyse the data and generate p values. These tell you whether to accept or reject the null hypothesis in the search for statistical significance. Ie are the results STATISTICALLY significant

24
Q

What is incidence?

A

Rate of occurrence new cases over period of time in a population

Measures risk of something.

(Eg mortality rate)

25
Q

What is prevalence?

A

Proportion of people with disease in given population at a point in time

Point prevalence

26
Q

What is qualitative data?

A

Categorical, non numerical eg outcome cured or not cured

27
Q

What is quantitative data?

A

Numerical. Can be discrete or continuous.

Discrete-finite number possible values eg number of people is 1 or 2 it can’t be 1.5.

Continuous data-infinite values eg weight.

Can turn it into categorical eg bp 80/40 is category hypotensive

28
Q

Nominal and ordinal data

A

Are categories that either have an inherent order eg mild moderate severe (ordinal) or none (nominal)

Analyse with non parametric statistics

29
Q

Interval and ratio scales to measure data….

A

Interval eg temperature in Celsius, points on scale have meaning. Can have zero not as a starting point just as a value.

Ratio scales sane as above but will start with a true zero

Analyse with parametric statistics

30
Q

Variance

A

Indicates dispersion of values around mean

Average distance by which each observation differs from the mean

Standard deviation describes spread around mean (variance is relating average distance of each observation from mean) square root of variance is one s.d

31
Q

What is a Z score?

A

Converts the value of an observation into the number of standard deviations that observation lies from the mean

32
Q

How to Interpret confidence interval?

A

If a confidence value comparing absolute difference between two groups includes zero it is statistically non significant

eg height difference is 3cm +/- 4cm between two groups

So difference could be between -1 and 7cm between groups.

Includes zero-zero difference =
Statistically non significant

If result expressed as a ratio between groups….eg relative risk or odds ratio the value of one is the value of no effect.

33
Q

What is per protocol analysis? Aka on treatment analysis?

A

When results of a trial include only those who completed trial as per protocol

May intro exclusion/attrition bias by excluding those lost to follow up, death, drop out. May often be those who failed to improve lost and may fail to demonstrate significant SE so may overestimate treatment effect

May not reflect real life

34
Q

What is intention to treat analysis?

A

All subjects randomised are included in analysis. Eg includes those who did not complete.

More closely mirrors real life.

35
Q

What is risk?

A

It means probability of something happening . Can be a number between 0 and 1 or a %

If no of people get ill is 1 in 6, risk is 1/6, 0.167, 16.7%

36
Q

What is odds?

A

A way to express chance.

Number of times an event is likely to occur compared to number of times not likely to occur. Eg if 1:6 fall ill, odds of falling ill is 1/5= 0.2

37
Q

What is a null hypothesis?

A

That there is no difference between two groups eg treatment and placebo

Use p value to say if results obtained in 2 groups are highly likely or unlikely due to chance

Allowing you to say with degree significance whether null hypothesis was correct or not ie there was indeed no difference….the smaller the p value the smaller the chance the null hypothesis is true.

38
Q

What are parametric tests?

A

Tests used to compare sets of (only) normally distributed data

Eg T test and x2 and ANOVA

39
Q

What is ANOVA?

A

Analysis of variance

Tells you whether two (or more) sample means come from same ‘population’ ie the null hypothesis ‘there is no difference between them’

T test is same but tests only 2 samples and tells you the probability the two come from a population with same mean value

40
Q

Mann Whitney test
Kruksal Wallis
Friedman
Wilcox on signed rank

A

Used to compare samples where mean is not helpful ie not normally distribution (non parametric)

Tests the hypothesis that there is no difference between the samples

If given median and Interquartile range can tell if parametric or not as would see normal (equal) distribution of data around median (which would also be the mean) if was parametric

41
Q

Chi2 X2 test

A

Test of significance look at the p value

Testing null hypothesis ‘there is no difference between sets of results’

Measures actual and expected frequencies

Can refine the test using

Yates
Fishers exact test

42
Q

What is risk?

A

Probability that an event will happen eg 1:100

43
Q

What is risk ratio?

A

Risk in one group compared to risk in another

Risk in exposed group/risk in control

Ratio of 1 indicates no difference

44
Q

What does RR (aka relative risk) mean if >1?

A

Mean risk in exposed is more than risk in controls

Usually given with CI
If CI does not cross 1 it is statistically significant

Used in cohort

45
Q

When is odds ratio used?

A

When looking for factors which do harm eg comparing cases to controls

46
Q

Clinical trial phases

A

0 micro dosing to elicit pharmacodynamics

1 healthy people (doses/SE)

2 illness in lab setting

3 illness in clinical setting

(Non inferiority)

MARKETING/AUTHORISATION

Post market surveillance data in diff populations/long term use

47
Q

Mann Whitney U test

A

Non parametric
Two unpaired samples
Median of one sample compared to median another

48
Q

Sign
Wilcoxon’s signed rank

A

One sample test
Non parametric
Median of one sample compared with a hypothetical mean

49
Q

Wilcoxons matched pairs

A

Paired sample
Non parametric
Medians compared in sample at different time points

50
Q

Kruksal Wallis anova

A

3+ non parametric samples
Medians compared

51
Q

Friedman’s

A

3+ paired samples

52
Q

McNemar

A

Paired categorical

53
Q

Fishers exact test

A

Unpaired categorical small sample

Alternative is use Yates correction on x2

MFSRH Part 2 2024 (43 decks)

Brainscape helps you reach your goals faster, through stronger study habits.
© 2024 Bold Learning Solutions. Terms and Conditions