The standard error of the mean (SEM) is a measure of the spread expected for the mean of the observations – i.e. how ‘accurate’ the calculated sample mean is from the true population mean. The relationship between the standard error of the mean and the standard deviation is such that, for a given sample size, the standard error of the mean equals the standard deviation divided by the square root of the sample size.

SEM = SD / square root (n)

where SD = standard deviation and n = sample size

Therefore, the SEM gets smaller as the sample size (n) increases.

If we want to depict how widely scattered some measurements are, we use the standard deviation. For indicating the uncertainty around the estimate of the mean, we use the standard error of the mean. The standard error is most useful as a means of calculating a confidence interval. For a large sample, a 95% confidence interval is obtained as the values 1.96×SE either side of the mean.

A 95% confidence interval:

lower limit = mean – (1.96 * SEM)

upper limit = mean + (1.96 * SEM)

Results such as mean value are often presented along with a confidence interval. For example, in a study the mean height in a sample taken from a population is 183cm. You know that the standard error (SE) (the standard deviation of the mean) is 2cm. This gives a 95% confidence interval of 179-187cm (+/- 2 SE).

Hence, it would be wrong to say that confidence levels do not apply to standard error of the mean.

The positive predictive value is the ratio of patients truly diagnosed as positive to all those who had positive test results. In this case, this is 60/(60+20)=75%.

The negative predictive value is the ratio of patients truly diagnosed as negative to all those who had negative test results. In this case, this is 80/(80+40)=67%.

The sensitivity is the ratio of patients with the disease who test positive i.e. true positive patients to the total number of people with the disease. In this case, this is 60/(60+40)=60%.

The specificity is the ratio of people who don’t have the disease who test negative i.e. true negatives to the total number of people without the disease. In this case, this is 80/(20+80)=80%.

70% is not the result of any screening measurements

Its known that 68.3% of the total values of a normally distributed variable are found within a range of 1 standard deviation from the mean which makes the range to be 4.3 to 4.9 mmol/L.

Normal distribution, also called Gaussian distribution, the most common distribution function for independent, randomly generated variables, and describes the spread for many biological and clinical measurements.

Properties of the Normal distribution

symmetrical i.e. Mean = mode = median

68.3% of values lie within 1 SD of the mean

95.4% of values lie within 2 SD of the mean

99.7% of values lie within 3 SD of the mean

The empirical rule, or the 68-95-99.7 rule, tells you where most of the values lie in a normal distribution: Around 68% of values are within 1 standard deviation of the mean.

Around 95% of values are within 2 standard deviations of the mean. Around 99.7% of values are within 3 standard deviations of the mean.
the standard deviation (SD) is a measure of how much dispersion exists from the mean.

SD = square root (variance)

The empirical rule, or the 68-95-99.7 rule states where most of the values lie in a normal distribution. Around 68% of values fall within 1 S.D of the mean, about 95% within 2 S.D of the mean, and about 99.7% of values within 3 S.D of the mean. Therefore, 95.4% is the most reasonable answer if results are normally distributed.

The degree of correlation is summarised by the correlation coefficient (r). This indicates how closely the points lie to a line drawn through the plotted data. In parametric data this is called Pearson’s correlation coefficient and can take any value between -1 to +1. A correlation of -1.0 indicates a perfect negative correlation, and a correlation of 1.0 indicates a perfect positive correlation.

For example

r = 1 – strong positive correlation (e.g. systolic blood pressure always increases with age)

r = 0 – no correlation (e.g. there is no correlation between systolic blood pressure and age)

r = – 1 – strong negative correlation (e.g. systolic blood pressure always decreases with age)

Whilst correlation coefficients give information about how one variable may increase or decrease as another variable increases they do not give information about how much the variable will change. They also do not provide information on cause and effect.

In contrast to the correlation coefficient, linear regression may be used to predict how much one variable changes when a second variable is changed.

Correlation doesn’t justify causality. Correlation coefficient gives us an idea whether or not the two parameters provide have any relation of some sort or not i.e. does change in one prompt any change in other? It has nothing to do with predictions. For that purpose linear regression is used.

Both incidence and prevalence are indicators of the disease frequency. While incidence tells us about the number of cases reported per population in a provided time period, prevalence is the factor you should be vigilant about as it tells us about the total number of cases that have been reported in a population at a particular point of time.

Mean, also known as the average, is the most common measure of central tendency. It is the sum of all observed values divided by the number of observation. It is not useful for skewed data, which has an abnormal distribution. It is useful, instead, for numerical data that have symmetric distribution. It reflects the contributions of each data in the group, and are sensitive to outliers.

The median is the value that falls in the middle position when the observations are ranked in order from the smallest to the largest. If the number of observations is odd, the median is the middle number. If it is even, the median is the average of the two middle numbers. Unlike the mean, the median is useful on skewed data, and can be used for ordinal or numerical data if skewed.

The mode is the value that occurs with the greatest frequency in a set of observations, and is utilized for bimodal distribution.

The variance and the standard deviation are not measures of central tendency, but of dispersion.

The normal distribution is a symmetrical, bell-shaped distribution in which the mean, median and mode are all equal.

9 & 27 can be obtained by subtracting and adding 9 from the mean. 9 is three times the standard deviation and we know that 99.7% values lie within 3 standard deviations from the mean. We can find the interval for 99.7% to verify in the following way:

For 99.7% confidence interval, you can find the range as follows:

1. Multiply the standard error by 3.

2. Subtract the answer from mean value to get the lower limit.

3. Add the answer obtained in step 1 from the mean value to get the upper limit.

4. The range turns out to be 9-27 kg.

This clinical trial consists over 1000 patients being evaluated for the response to a new treatment against a currently licensed treatment for ulcerative colitis. Therefore, it is comparing its efficacy to an established therapeutic or control in a larger population of volunteers. These are the characteristics of a phase III clinical trial.

Being the part of a meta analysis, forest plots are more valued as evidence then randomised control trials.

The notion that forest plots can only be used if the results are substantial is not true. They are good indicators of the significance of the data. If the diamond intersects the central line, the data is rendered significant. It also aggregates means and confidence intervals from studies conducted in the past which makes the study much more reliable as errors associated with individual studies tend to have less of an impact in this way.

The suggestion that forest plots are primarily used for qualitative data is factually incorrect. Forest plots require numerical values to function.

All in all, forest plots help us in determining whether or not there is a significant trend in that particular field of study.

In statistical terms, correlation is used to denote association between two quantitative variables.

The degree of association is measured by a correlation coefficient, denoted by r. The correlation coefficient is measured on a scale that varies from + 1 through 0 to – 1. Complete correlation between two variables is expressed by either + 1 or -1. When one variable increases as the other increases the correlation is positive; when one decreases as the other increases it is negative. Complete absence of correlation is represented by 0.

The two methods are not synonymous as correlation measures the degree of relationship between two variables whereas regression analysis is about how one variable affects another or what changes it has on the other variable. Both are also shown by a different graphical representation.

A prospective randomised double blind controlled trial against cyclizine in multiple centres is the most likely to change your practice.

Case controlled studies are efficient in identifying an association between a drug treatment and outcome and are usually conducted retrospectively. They are generally less valued than prospective randomised trials. They cannot generate incidence data, are subject to bias, have difficult selection of controls and can be made more difficult if note keeping is not reliable.

The gold standard in intervention-based studies is randomised controlled double blind trials. Its features are:

Treating all intervention groups identically
Reduction of bias by random allocation to intervention groups
Patients and researchers unaware of which treatment was given until at completion of study
Patients analysed within the group to which they were allocated, and
Analysis focused on estimating the size of the difference in predefined outcomes between intervention groups.

New healthcare interventions should be evaluated through properly designed randomised controlled trials (though there are some potential ethical disadvantages)

Conducting trials in multiple centres is an accepted way of evaluating a new drug as it may be the only way of recruiting sufficient number of patients within a reasonable time frame to satisfy the objectives of the trial. Type II statistical errors will occur if a small numbers of patients is used in study group.

The hazard ratio (HR) is simply a comparison of two hazards in a study. It provides an estimate of the ratio of the hazard rates between the experimental group and a control group over the entire study duration. It is typically used when analysing survival over time, hence is the most suitable statistical method in this case.

An odds ratio is a statistic that quantifies the strength of the association between two events, A and B. It is the “measure of association” for a case-control study.

The Pearson product-moment correlation coefficient (Pearson’s correlation, for short) is a measure of the strength and direction of association that exists between two variables. An example would be if scientists wanted to evaluate the relationship between quality of certain population of rice and their genetic make-up.

Relative risk is the ratio of the risks for an event for the exposure group to the risks for the non-exposure group. Thus relative risk provides an increase or decrease in the likelihood of an event based on some exposure. Relative risk measures the association between the exposure and the outcome.

Absolute risk reduction is the number of percentage points your own risk goes down if you do a preventive act such as stop drinking alcohol. It depends on what your risk factors are to begin with.

Observers may subconsciously measure or report data in a way that favours the expected study outcome. Therefore, by blinding the study we can eliminate expectation bias.

Recall bias is a systematic error that occurs when the study participants omit details or do not remember previous events or experiences accurately.

Verification can occur during investigations when there is a difference in testing strategy between groups of individuals, which might lead to biasness due to differing ways of verifying the disease of interest.

Nonresponse bias is the bias that occurs when the people who respond to a survey differ significantly from the people who do not respond to the survey.

A distortion that modifies an association between an exposure and an outcome because a factor is independently associated with the exposure and the outcome. Randomization is the best way to reduce the risk of confounding.

Work up bias involves comparing the novel diagnostic test with the current standard test. A portion of the patients undergo the standard test while others undergo the new test as the standard test is costly. The result can be alteration in specify and sensitivity.

Recall bias is specifically appropriate to the case control studies that is when ever the memories retrieved by the participants differ in accuracy.

Not publishing the results of a valid study just because they are negative or uninteresting can be termed as publication bias.

When information gathering is ill suited with respect to time i.e. collecting the data regarding a fatal disease many years after the death of its patients, it is termed as Late – look bias.

The case in point is an instance of lead time bias when upon comparing two tests, one is able to detect the condition earlier than the other but the overall outcome doesn’t change. There is a possibility that this will make the survival rates for the newer test look more promising.

68% of the population will be found in one standard deviation (SD) above plus one SD below the mean. Two SDs above plus two SDs below the mean will include 95% of the population.

The median can be greater or less than the mean as it is simply the mid point of the data after the data is arranged. Half the data are above and half below the median .

The mode is a true score, unlike the mean or the median. It is the most common score or the score obtained from the largest number of subjects in any given data.

The dose-response curve plots the graph of the dose (drug concentration) versus the response. As doses increase, the response increment diminishes; finally, doses may be reached at which no further increase in response can be achieved. This relation between drug concentration and effect is traditionally described by a hyperbolic curve. When the x-axis is plotted in log scale, the graph yields a sigmoid curve.

Efficacy (Emax) and potency (EC50) can be derived from this curve. Emax is the maximal effect achievable, with increasing concentration of a drug. EC50 is the concentration of the drug, wherein half of the maximal effect is achieved.

When the graph is plotted using a log [response/1-response] against log dose, the sigmoid curve becomes a straight line (Hill plot). A graph that transforms from a straight line to exponential curve is mathematically incorrect. A graph that transforms from either a wash-in or wash-out exponential curve to a straight line comes from an initial set of data plotted against time, to a logarithmic transformation of the initial data set against time.

ANOVA is based upon within group variance (i.e. the variance of the mean of a sample) and between group variance (i.e. the variance between means of different samples). The test works by finding out the ratio of the two variances mentioned above. (Commonly known as F statistic).

Hawthorne effect explains the change in any behavioural aspect owing to the awareness that the person is being observed.
Simpson’s Paradox explains the association developed when the data from several groups is combined to form a single larger group.

The remaining terms are made up.

Observation bias occurs when the behaviour of an individual changes that results from their awareness of being observed.

Recall bias introduced when participants in a study are systematically more or less likely to recall and relate information on exposure depending on their outcome status.

Attrition bias is a systematic error caused by unequal loss of participants from a randomized controlled trial (RCT). In clinical trials, participants might dropout due to unsatisfactory treatment or efficacy, intolerable adverse events, or even death.

Selection bias introduced when the individuals are not chosen randomly to take a part in the study. It usually occurs when the research decides who is going to be studied, they are not the representative of the population.

Lead-time bias occurs when a disease is detected by a screening test at an earlier time point rather than it would have been diagnosed by its clinical appearance. In this bias, earlier detection improves the survival time in the intervention group.

ANOVA is based upon within group variance (i.e. the variance of the mean of a sample) and between group variance (i.e. the variance between means of different samples). The test works by finding out the ratio of the two variances mentioned above. (Commonly known as F statistic).

The Delphi technique was developed in the 1950s and is a widely used and accepted method for achieving convergence of opinion concerning real-world knowledge solicited from experts within certain topic areas. Choosing the appropriate subjects is the most important step in the entire process because it directly relates to the quality of the results generated, despite this, there is no exact criterion currently listed in the literature concerning the selection of Delphi participants.

Therefore, due to the potential scarcity of qualified participants and the relatively small number of subjects used in a Delphi study, the ability to achieve and maintain an ideal response rate can either ensure or jeopardize the validity of a Delphi study.

99.7% of the values within 3 standard deviations of the mean.

For 99.7% confidence interval, you can find the range as follows:

1. Multiply the standard error by 3.

2. Subtract the answer from mean value to get the lower limit.

3. Add the answer obtained in step 1 from the mean value to get the upper limit.

For a confidence interval of 68%, multiply the standard error with 1 and repeat the process. For a 95% confidence interval, Standard Error is multiplied by 1.96 to get the interval.

Linear regression, using a technique called curve fitting, allows us to make predictions regarding a certain variable.

Correlation coefficient gives us an idea whether or not the two parameters provide have any relation of some sort or not i.e. does change in one prompt any change in other?

Absolute risk reduction = (Control event rate) – (Experimental event rate)

Experimental event rate = 24 / 120 = 0.2

Control event rate = 60 / 240 = 0.25

Absolute risk reduction = 0.25 – 0.2 = 0.05 = 5% reduction

During analytical stage a technique called stratification is used for controlling confounding variables. This technique involves sorting out the data into discernible groups.

Standard error can be calculated by the following formula:
Standard Error= (Standard Deviation)/√(Sample Size)
= (16) / √(100)
= 16 / 10
= 1.6

Level 1 – High-quality randomised controlled trial with statistically significant difference or no statistically significant difference but narrow confidence intervals (prospective controlled)

Level 2 – Prospective comparative study (prospective uncontrolled)

Level 3 – Case-control study, retrospective comparative study (retrospective controlled)

Level 4 – Case series (retrospective uncontrolled)

Level 5 – Expert opinion.