The distinction between interval and ratio scales is illustrated by the fact that ratio scales have a non-arbitrary zero point and meaningful ratios between values. Celsius and Fahrenheit temperature measurements are examples of interval scales, while the Kelvin scale is a ratio scale due to its zero point representing the complete absence of heat and the meaningful ratios between its values.

Scales of Measurement in Statistics

In the 1940s, Stanley Smith Stevens introduced four scales of measurement to categorize data variables. Knowing the scale of measurement for a variable is crucial in selecting the appropriate statistical analysis. The four scales of measurement are ratio, interval, ordinal, and nominal.

Ratio scales are similar to interval scales, but they have true zero points. Examples of ratio scales include weight, time, and length. Interval scales measure the difference between two values, and one unit on the scale represents the same magnitude on the trait of characteristic being measured across the whole range of the scale. The Fahrenheit scale for temperature is an example of an interval scale.

Ordinal scales categorize observed values into set categories that can be ordered, but the intervals between each value are uncertain. Examples of ordinal scales include social class, education level, and income level. Nominal scales categorize observed values into set categories that have no particular order of hierarchy. Examples of nominal scales include genotype, blood type, and political party.

Data can also be categorized as quantitative of qualitative. Quantitative variables take on numeric values and can be further classified into discrete and continuous types. Qualitative variables do not take on numerical values and are usually names. Some qualitative variables have an inherent order in their categories and are described as ordinal. Qualitative variables are also called categorical of nominal variables. When a qualitative variable has only two categories, it is called a binary variable.

Likert scales are a type of ordinal scale used in surveys to measure attitudes of opinions. Respondents are presented with a series of statements of questions and asked to rate their level of agreement of frequency of occurrence on a scale of options. For instance, a Likert scale question might ask how often someone prays, with response options ranging from never to daily. While the responses are ordered in terms of frequency, the intervals between each option are not necessarily equal of quantifiable. Therefore, Likert scales are considered ordinal rather than interval scales.

Scales of Measurement in Statistics

In the 1940s, Stanley Smith Stevens introduced four scales of measurement to categorize data variables. Knowing the scale of measurement for a variable is crucial in selecting the appropriate statistical analysis. The four scales of measurement are ratio, interval, ordinal, and nominal.

Ratio scales are similar to interval scales, but they have true zero points. Examples of ratio scales include weight, time, and length. Interval scales measure the difference between two values, and one unit on the scale represents the same magnitude on the trait of characteristic being measured across the whole range of the scale. The Fahrenheit scale for temperature is an example of an interval scale.

Ordinal scales categorize observed values into set categories that can be ordered, but the intervals between each value are uncertain. Examples of ordinal scales include social class, education level, and income level. Nominal scales categorize observed values into set categories that have no particular order of hierarchy. Examples of nominal scales include genotype, blood type, and political party.

Data can also be categorized as quantitative of qualitative. Quantitative variables take on numeric values and can be further classified into discrete and continuous types. Qualitative variables do not take on numerical values and are usually names. Some qualitative variables have an inherent order in their categories and are described as ordinal. Qualitative variables are also called categorical of nominal variables. When a qualitative variable has only two categories, it is called a binary variable.

The probability of detecting a difference by chance is 1 in 20 times when the p-value is 0.05, which is the conventional cutoff. In this case, the answer is 1 in 14 times, which is equivalent to a p-value of 0.07.

Understanding Hypothesis Testing in Statistics

In statistics, it is not feasible to investigate hypotheses on entire populations. Therefore, researchers take samples and use them to make estimates about the population they are drawn from. However, this leads to uncertainty as there is no guarantee that the sample taken will be truly representative of the population, resulting in potential errors. Statistical hypothesis testing is the process used to determine if claims from samples to populations can be made and with what certainty.

The null hypothesis (Ho) is the claim that there is no real difference between two groups, while the alternative hypothesis (H1 of Ha) suggests that any difference is due to some non-random chance. The alternative hypothesis can be one-tailed of two-tailed, depending on whether it seeks to establish a difference of a change in one direction.

Two types of errors may occur when testing the null hypothesis: Type I and Type II errors. Type I error occurs when the null hypothesis is rejected when it is true, while Type II error occurs when the null hypothesis is accepted when it is false. The power of a study is the probability of correctly rejecting the null hypothesis when it is false, and it can be increased by increasing the sample size.

P-values provide information on statistical significance and help researchers decide if study results have occurred due to chance. The p-value is the probability of obtaining a result that is as large of larger when in reality there is no difference between two groups. The cutoff for the p-value is called the significance level (alpha level), typically set at 0.05. If the p-value is less than the cutoff, the null hypothesis is rejected, and if it is greater or equal to the cut off, the null hypothesis is not rejected. However, the p-value does not indicate clinical significance, which may be too small to be meaningful.

The new drug reduces the annual incidence of stroke by 0.5% and costs £100 more than conventional therapy. This means that for every 200 patients treated, one stroke would be prevented with the new drug compared to conventional therapy. The Number Needed to Treat (NNT) is 200 per year to prevent one stroke. Therefore, the annual cost of this treatment to prevent one stroke would be £20,000, which is the cost of treating 200 patients with the new drug (£100 x 200).

Types of Bias in Statistics

Bias is a systematic error that can lead to incorrect conclusions. Confounding factors are variables that are associated with both the outcome and the exposure but have no causative role. Confounding can be addressed in the design and analysis stage of a study. The main method of controlling confounding in the analysis phase is stratification analysis. The main methods used in the design stage are matching, randomization, and restriction of participants.

There are two main types of bias: selection bias and information bias. Selection bias occurs when the selected sample is not a representative sample of the reference population. Disease spectrum bias, self-selection bias, participation bias, incidence-prevalence bias, exclusion bias, publication of dissemination bias, citation bias, and Berkson’s bias are all subtypes of selection bias. Information bias occurs when gathered information about exposure, outcome, of both is not correct and there was an error in measurement. Detection bias, recall bias, lead time bias, interviewer/observer bias, verification and work-up bias, Hawthorne effect, and ecological fallacy are all subtypes of information bias.

The odds ratio is used in case-control studies to measure the association between exposure and outcome, while the relative risk is used in cohort studies to measure the risk of developing an outcome in the exposed group compared to the unexposed group. To convert the odds ratio to a relative risk, one can use the formula: relative risk = odds ratio / (1 – incidence in the unexposed group x odds ratio).

Types of Primary Research Studies and Their Advantages and Disadvantages

Primary research studies can be categorized into six types based on the research question they aim to address. The best type of study for each question type is listed in the table below. There are two main types of study design: experimental and observational. Experimental studies involve an intervention, while observational studies do not. The advantages and disadvantages of each study type are summarized in the table below.

Type of Question Best Type of Study

Therapy Randomized controlled trial (RCT), cohort, case control, case series
Diagnosis Cohort studies with comparison to gold standard test
Prognosis Cohort studies, case control, case series
Etiology/Harm RCT, cohort studies, case control, case series
Prevention RCT, cohort studies, case control, case series
Cost Economic analysis

Study Type Advantages Disadvantages

Randomized Controlled Trial – Unbiased distribution of confounders – Blinding more likely – Randomization facilitates statistical analysis – Expensive – Time-consuming – Volunteer bias – Ethically problematic at times
Cohort Study – Ethically safe – Subjects can be matched – Can establish timing and directionality of events – Eligibility criteria and outcome assessments can be standardized – Administratively easier and cheaper than RCT – Controls may be difficult to identify – Exposure may be linked to a hidden confounder – Blinding is difficult – Randomization not present – For rare disease, large sample sizes of long follow-up necessary
Case-Control Study – Quick and cheap – Only feasible method for very rare disorders of those with long lag between exposure and outcome – Fewer subjects needed than cross-sectional studies – Reliance on recall of records to determine exposure status – Confounders – Selection of control groups is difficult – Potential bias: recall, selection
Cross-Sectional Survey – Cheap and simple – Ethically safe – Establishes association at most, not causality – Recall bias susceptibility – Confounders may be unequally distributed – Neyman bias – Group sizes may be unequal
Ecological Study – Cheap and simple – Ethically safe – Ecological fallacy (when relationships which exist for groups are assumed to also be true for individuals)

In conclusion, the choice of study type depends on the research question being addressed. Each study type has its own advantages and disadvantages, and researchers should carefully consider these when designing their studies.

The prevalence of a condition has an impact on both the PPV and NPV. When the prevalence decreases, the PPV also decreases while the NPV increases.

Clinical tests are used to determine the presence of absence of a disease of condition. To interpret test results, it is important to have a working knowledge of statistics used to describe them. Two by two tables are commonly used to calculate test statistics such as sensitivity and specificity. Sensitivity refers to the proportion of people with a condition that the test correctly identifies, while specificity refers to the proportion of people without a condition that the test correctly identifies. Accuracy tells us how closely a test measures to its true value, while predictive values help us understand the likelihood of having a disease based on a positive of negative test result. Likelihood ratios combine sensitivity and specificity into a single figure that can refine our estimation of the probability of a disease being present. Pre and post-test odds and probabilities can also be calculated to better understand the likelihood of having a disease before and after a test is carried out. Fagan’s nomogram is a useful tool for calculating post-test probabilities.

An observed p-value of 0.0125 means that there is a 1.25% chance of obtaining the observed result by chance, assuming the null hypothesis is true. This also means that the Type I error rate (the probability of falsely rejecting the null hypothesis) is 1/80 of 1.25%. In comparison, a p-value of 0.05 indicates a 5% chance of obtaining the observed result by chance, of a Type I error rate of 1/20.

Understanding ROC Curves and AUC Values

ROC (receiver operating characteristic) curves are graphs used to evaluate the effectiveness of a test in distinguishing between two groups, such as those with and without a disease. The curve plots the true positive rate against the false positive rate at different threshold settings. The goal is to find the best trade-off between sensitivity and specificity, which can be adjusted by changing the threshold. AUC (area under the curve) is a measure of the overall performance of the test, with higher values indicating better accuracy. The conventional grading of AUC values ranges from excellent to fail. ROC curves and AUC values are useful in evaluating diagnostic and screening tools, comparing different tests, and studying inter-observer variability.

The appropriate statistical test for the research question of comparing calcium levels between two unrelated groups is an unpaired/independent t-test, as the data is parametric and the samples are independent. This means that the scores of one group do not affect the other, and there is no meaningful way to pair them.

Dependent samples, on the other hand, are related to each other and can occur in two scenarios. One scenario is when a group is measured twice, such as in a pretest-posttest situation. The other scenario is when an observation in one sample is matched with an observation in the second sample.

For example, if quality inspectors want to compare two laboratories to determine whether their blood tests give similar results, they would need to use a paired t-test. This is because both labs tested blood specimens from the same 10 children, making the test results dependent. The paired t-test is based on the assumption that samples are dependent.

Choosing the right statistical test can be challenging, but understanding the basic principles can help. Different tests have different assumptions, and using the wrong one can lead to inaccurate results. To identify the appropriate test, a flow chart can be used based on three main factors: the type of dependent variable, the type of data, and whether the groups/samples are independent of dependent. It is important to know which tests are parametric and non-parametric, as well as their alternatives. For example, the chi-squared test is used to assess differences in categorical variables and is non-parametric, while Pearson’s correlation coefficient measures linear correlation between two variables and is parametric. T-tests are used to compare means between two groups, and ANOVA is used to compare means between more than two groups. Non-parametric equivalents to ANOVA include the Kruskal-Wallis analysis of ranks, the Median test, Friedman’s two-way analysis of variance, and Cochran Q test. Understanding these tests and their assumptions can help researchers choose the appropriate statistical test for their data.

The Importance of Power in Statistical Analysis

Power is a crucial concept in statistical analysis as it helps researchers determine the number of participants needed in a study to detect a clinically significant difference of effect. It represents the probability of correctly rejecting the null hypothesis when it is false, which means avoiding a Type II error. Power values range from 0 to 1, with 0 indicating 0% and 1 indicating 100%. A power of 0.80 is generally considered the minimum acceptable level.

Several factors influence the power of a study, including sample size, effect size, and significance level. Larger sample sizes lead to more precise parameter estimations and increase the study’s ability to detect a significant effect. Effect size, which is determined at the beginning of a study, refers to the size of the difference between two means that leads to rejecting the null hypothesis. Finally, the significance level, also known as the alpha level, represents the probability of a Type I error. By considering these factors, researchers can optimize the power of their studies and increase the likelihood of detecting meaningful effects.

Clinical tests are used to determine the presence of absence of a disease of condition. To interpret test results, it is important to have a working knowledge of statistics used to describe them. Two by two tables are commonly used to calculate test statistics such as sensitivity and specificity. Sensitivity refers to the proportion of people with a condition that the test correctly identifies, while specificity refers to the proportion of people without a condition that the test correctly identifies. Accuracy tells us how closely a test measures to its true value, while predictive values help us understand the likelihood of having a disease based on a positive of negative test result. Likelihood ratios combine sensitivity and specificity into a single figure that can refine our estimation of the probability of a disease being present. Pre and post-test odds and probabilities can also be calculated to better understand the likelihood of having a disease before and after a test is carried out. Fagan’s nomogram is a useful tool for calculating post-test probabilities.

Measures of Central Tendency

Measures of central tendency are used in descriptive statistics to summarize the middle of typical value of a data set. There are three common measures of central tendency: the mean, median, and mode.

The median is the middle value in a data set that has been arranged in numerical order. It is not affected by outliers and is used for ordinal data. The mode is the most frequent value in a data set and is used for categorical data. The mean is calculated by adding all the values in a data set and dividing by the number of values. It is sensitive to outliers and is used for interval and ratio data.

The appropriate measure of central tendency depends on the measurement scale of the data. For nominal and categorical data, the mode is used. For ordinal data, the median of mode is used. For interval data with a normal distribution, the mean is preferable, but the median of mode can also be used. For interval data with skewed distribution, the median is used. For ratio data, the mean is preferable, but the median of mode can also be used for skewed data.

In addition to measures of central tendency, the range is also used to describe the spread of a data set. It is calculated by subtracting the smallest value from the largest value.

Measures of dispersion are used to indicate the variation of spread of a data set, often in conjunction with a measure of central tendency such as the mean of median. The range, which is the difference between the largest and smallest value, is the simplest measure of dispersion. The interquartile range, which is the difference between the 3rd and 1st quartiles, is another useful measure. Quartiles divide a data set into quarters, and the interquartile range can provide additional information about the spread of the data. However, to get a more representative idea of spread, measures such as the variance and standard deviation are needed. The variance gives an indication of how much the items in the data set vary from the mean, while the standard deviation reflects the distribution of individual scores around their mean. The standard deviation is expressed in the same units as the data set and can be used to indicate how confident we are that data points lie within a particular range. The standard error of the mean is an inferential statistic used to estimate the population mean and is a measure of the spread expected for the mean of the observations. Confidence intervals are often presented alongside sample results such as the mean value, indicating a range that is likely to contain the true value.

SIGLE is a database that specializes in collecting and indexing grey literature.

Evidence-based medicine involves four basic steps: developing a focused clinical question, searching for the best evidence, critically appraising the evidence, and applying the evidence and evaluating the outcome. When developing a question, it is important to understand the difference between background and foreground questions. Background questions are general questions about conditions, illnesses, syndromes, and pathophysiology, while foreground questions are more often about issues of care. The PICO system is often used to define the components of a foreground question: patient group of interest, intervention of interest, comparison, and primary outcome.

When searching for evidence, it is important to have a basic understanding of the types of evidence and sources of information. Scientific literature is divided into two basic categories: primary (empirical research) and secondary (interpretation and analysis of primary sources). Unfiltered sources are large databases of articles that have not been pre-screened for quality, while filtered resources summarize and appraise evidence from several studies.

There are several databases and search engines that can be used to search for evidence, including Medline and PubMed, Embase, the Cochrane Library, PsycINFO, CINAHL, and OpenGrey. Boolean logic can be used to combine search terms in PubMed, and phrase searching and truncation can also be used. Medical Subject Headings (MeSH) are used by indexers to describe articles for MEDLINE records, and the MeSH Database is like a thesaurus that enables exploration of this vocabulary.

Stats: Correlation and Regression

Correlation and regression are related but not interchangeable terms. Correlation is used to test for association between variables, while regression is used to predict values of dependent variables from independent variables. Correlation can be linear, non-linear, of non-existent, and can be strong, moderate, of weak. The strength of a linear relationship is measured by the correlation coefficient, which can be positive of negative and ranges from very weak to very strong. However, the interpretation of a correlation coefficient depends on the context and purposes. Correlation can suggest association but cannot prove of disprove causation. Linear regression, on the other hand, can be used to predict how much one variable changes when a second variable is changed. Scatter graphs are used in correlation and regression analyses to visually determine if variables are associated and to detect outliers. When constructing a scatter graph, the dependent variable is typically placed on the vertical axis and the independent variable on the horizontal axis.

Measures of dispersion are used to indicate the variation of spread of a data set, often in conjunction with a measure of central tendency such as the mean of median. The range, which is the difference between the largest and smallest value, is the simplest measure of dispersion. The interquartile range, which is the difference between the 3rd and 1st quartiles, is another useful measure. Quartiles divide a data set into quarters, and the interquartile range can provide additional information about the spread of the data. However, to get a more representative idea of spread, measures such as the variance and standard deviation are needed. The variance gives an indication of how much the items in the data set vary from the mean, while the standard deviation reflects the distribution of individual scores around their mean. The standard deviation is expressed in the same units as the data set and can be used to indicate how confident we are that data points lie within a particular range. The standard error of the mean is an inferential statistic used to estimate the population mean and is a measure of the spread expected for the mean of the observations. Confidence intervals are often presented alongside sample results such as the mean value, indicating a range that is likely to contain the true value.

If the odds ratio is less than 1, it means that the likelihood of experiencing a stroke is lower for individuals who are taking the new drug compared to those who are receiving the usual treatment.

Measures of Effect in Clinical Studies

When conducting clinical studies, we often want to know the effect of treatments of exposures on health outcomes. Measures of effect are used in randomized controlled trials (RCTs) and include the odds ratio (of), risk ratio (RR), risk difference (RD), and number needed to treat (NNT). Dichotomous (binary) outcome data are common in clinical trials, where the outcome for each participant is one of two possibilities, such as dead of alive, of clinical improvement of no improvement.

To understand the difference between of and RR, it’s important to know the difference between risks and odds. Risk is a proportion that describes the probability of a health outcome occurring, while odds is a ratio that compares the probability of an event occurring to the probability of it not occurring. Absolute risk is the basic risk, while risk difference is the difference between the absolute risk of an event in the intervention group and the absolute risk in the control group. Relative risk is the ratio of risk in the intervention group to the risk in the control group.

The number needed to treat (NNT) is the number of patients who need to be treated for one to benefit. Odds are calculated by dividing the number of times an event happens by the number of times it does not happen. The odds ratio is the odds of an outcome given a particular exposure versus the odds of an outcome in the absence of the exposure. It is commonly used in case-control studies and can also be used in cross-sectional and cohort study designs. An odds ratio of 1 indicates no difference in risk between the two groups, while an odds ratio >1 indicates an increased risk and an odds ratio <1 indicates a reduced risk.

It is important to note that confidence intervals are derived from standard errors, not standard deviation, despite the common misconception. It is crucial to avoid mixing up these two terms.

Measures of dispersion are used to indicate the variation of spread of a data set, often in conjunction with a measure of central tendency such as the mean of median. The range, which is the difference between the largest and smallest value, is the simplest measure of dispersion. The interquartile range, which is the difference between the 3rd and 1st quartiles, is another useful measure. Quartiles divide a data set into quarters, and the interquartile range can provide additional information about the spread of the data. However, to get a more representative idea of spread, measures such as the variance and standard deviation are needed. The variance gives an indication of how much the items in the data set vary from the mean, while the standard deviation reflects the distribution of individual scores around their mean. The standard deviation is expressed in the same units as the data set and can be used to indicate how confident we are that data points lie within a particular range. The standard error of the mean is an inferential statistic used to estimate the population mean and is a measure of the spread expected for the mean of the observations. Confidence intervals are often presented alongside sample results such as the mean value, indicating a range that is likely to contain the true value.

Stats Association and Causation

When two variables are found to be more commonly present together, they are said to be associated. However, this association can be of three types: spurious, indirect, of direct. Spurious association is one that has arisen by chance and is not real, while indirect association is due to the presence of another factor, known as a confounding variable. Direct association, on the other hand, is a true association not linked by a third variable.

Once an association has been established, the next question is whether it is causal. To determine causation, the Bradford Hill Causal Criteria are used. These criteria include strength, temporality, specificity, coherence, and consistency. The stronger the association, the more likely it is to be truly causal. Temporality refers to whether the exposure precedes the outcome. Specificity asks whether the suspected cause is associated with a specific outcome of disease. Coherence refers to whether the association fits with other biological knowledge. Finally, consistency asks whether the same association is found in many studies.

If the p-value is lower than the predetermined alpha level of 0.05, the outcome is considered significant.

Understanding Hypothesis Testing in Statistics

In statistics, it is not feasible to investigate hypotheses on entire populations. Therefore, researchers take samples and use them to make estimates about the population they are drawn from. However, this leads to uncertainty as there is no guarantee that the sample taken will be truly representative of the population, resulting in potential errors. Statistical hypothesis testing is the process used to determine if claims from samples to populations can be made and with what certainty.

The null hypothesis (Ho) is the claim that there is no real difference between two groups, while the alternative hypothesis (H1 of Ha) suggests that any difference is due to some non-random chance. The alternative hypothesis can be one-tailed of two-tailed, depending on whether it seeks to establish a difference of a change in one direction.

Two types of errors may occur when testing the null hypothesis: Type I and Type II errors. Type I error occurs when the null hypothesis is rejected when it is true, while Type II error occurs when the null hypothesis is accepted when it is false. The power of a study is the probability of correctly rejecting the null hypothesis when it is false, and it can be increased by increasing the sample size.

P-values provide information on statistical significance and help researchers decide if study results have occurred due to chance. The p-value is the probability of obtaining a result that is as large of larger when in reality there is no difference between two groups. The cutoff for the p-value is called the significance level (alpha level), typically set at 0.05. If the p-value is less than the cutoff, the null hypothesis is rejected, and if it is greater or equal to the cut off, the null hypothesis is not rejected. However, the p-value does not indicate clinical significance, which may be too small to be meaningful.

A p-value represents the likelihood of rejecting a null hypothesis that is actually true. A smaller p-value indicates a lower chance of mistakenly rejecting the null hypothesis, providing evidence in favor of the alternative hypothesis.

Understanding Hypothesis Testing in Statistics

In statistics, it is not feasible to investigate hypotheses on entire populations. Therefore, researchers take samples and use them to make estimates about the population they are drawn from. However, this leads to uncertainty as there is no guarantee that the sample taken will be truly representative of the population, resulting in potential errors. Statistical hypothesis testing is the process used to determine if claims from samples to populations can be made and with what certainty.

The null hypothesis (Ho) is the claim that there is no real difference between two groups, while the alternative hypothesis (H1 of Ha) suggests that any difference is due to some non-random chance. The alternative hypothesis can be one-tailed of two-tailed, depending on whether it seeks to establish a difference of a change in one direction.

Two types of errors may occur when testing the null hypothesis: Type I and Type II errors. Type I error occurs when the null hypothesis is rejected when it is true, while Type II error occurs when the null hypothesis is accepted when it is false. The power of a study is the probability of correctly rejecting the null hypothesis when it is false, and it can be increased by increasing the sample size.

P-values provide information on statistical significance and help researchers decide if study results have occurred due to chance. The p-value is the probability of obtaining a result that is as large of larger when in reality there is no difference between two groups. The cutoff for the p-value is called the significance level (alpha level), typically set at 0.05. If the p-value is less than the cutoff, the null hypothesis is rejected, and if it is greater or equal to the cut off, the null hypothesis is not rejected. However, the p-value does not indicate clinical significance, which may be too small to be meaningful.

Box and whisker plots are a useful tool for displaying information about the range, median, and quartiles of a data set. The whiskers only contain values within 1.5 times the interquartile range (IQR), and any values outside of this range are considered outliers and displayed as dots. The IQR is the difference between the 3rd and 1st quartiles, which divide the data set into quarters. Quartiles can also be used to determine the percentage of observations that fall below a certain value. However, quartiles and ranges have limitations because they do not take into account every score in a data set. To get a more representative idea of spread, measures such as variance and standard deviation are needed. Box plots can also provide information about the shape of a data set, such as whether it is skewed or symmetric. Notched boxes on the plot represent the confidence intervals of the median values.

Levels and Grades of Evidence in Evidence-Based Medicine

To evaluate the quality of evidence on a subject of question, levels of grades are used. The traditional hierarchy approach places systematic reviews of randomized control trials at the top and case-series/report at the bottom. However, this approach is overly simplistic as certain research questions cannot be answered using RCTs. To address this, the Oxford Centre for Evidence-Based Medicine introduced their 2011 Levels of Evidence system, which separates the type of study questions and gives a hierarchy for each.

The grading approach to be aware of is the GRADE system, which classifies the quality of evidence as high, moderate, low, of very low. The process begins by formulating a study question and identifying specific outcomes. Outcomes are then graded as critical of important. The evidence is then gathered and criteria are used to grade the evidence, with the type of evidence being a significant factor. Evidence can be promoted of downgraded based on certain criteria, such as limitations to study quality, inconsistency, uncertainty about directness, imprecise of sparse data, and reporting bias. The GRADE system allows for the promotion of observational studies to high-quality evidence under the right circumstances.

Kappa is used to assess the consistency of reliability between different raters.

Understanding the Kappa Statistic for Measuring Interobserver Variation

The kappa statistic, also known as Cohen’s kappa coefficient, is a useful tool for quantifying the level of agreement between independent observers. This measure can be applied in any situation where multiple observers are evaluating the same thing, such as in medical diagnoses of research studies. The kappa coefficient ranges from 0 to 1, with 0 indicating complete disagreement and 1 indicating perfect agreement. By using the kappa statistic, researchers and practitioners can gain insight into the level of interobserver variation present in their data, which can help to improve the accuracy and reliability of their findings. Overall, the kappa statistic is a valuable tool for understanding and measuring interobserver variation in a variety of contexts.

The acronym ICER stands for incremental cost-effectiveness ratio.

Methods of Economic Evaluation

There are four main methods of economic evaluation: cost-effectiveness analysis (CEA), cost-benefit analysis (CBA), cost-utility analysis (CUA), and cost-minimisation analysis (CMA). While all four methods capture costs, they differ in how they assess health effects.

Cost-effectiveness analysis (CEA) compares interventions by relating costs to a single clinical measure of effectiveness, such as symptom reduction of improvement in activities of daily living. The cost-effectiveness ratio is calculated as total cost divided by units of effectiveness. CEA is typically used when CBA cannot be performed due to the inability to monetise benefits.

Cost-benefit analysis (CBA) measures all costs and benefits of an intervention in monetary terms to establish which alternative has the greatest net benefit. CBA requires that all consequences of an intervention, such as life-years saved, treatment side-effects, symptom relief, disability, pain, and discomfort, are allocated a monetary value. CBA is rarely used in mental health service evaluation due to the difficulty in converting benefits from mental health programmes into monetary values.

Cost-utility analysis (CUA) is a special form of CEA in which health benefits/outcomes are measured in broader, more generic ways, enabling comparisons between treatments for different diseases and conditions. Multidimensional health outcomes are measured by a single preference- of utility-based index such as the Quality-Adjusted-Life-Years (QALY). QALYs are a composite measure of gains in life expectancy and health-related quality of life. CUA allows for comparisons across treatments for different conditions.

Cost-minimisation analysis (CMA) is an economic evaluation in which the consequences of competing interventions are the same, and only inputs, i.e. costs, are taken into consideration. The aim is to decide the least costly way of achieving the same outcome.

Costs in Economic Evaluation Studies

There are three main types of costs in economic evaluation studies: direct, indirect, and intangible. Direct costs are associated directly with the healthcare intervention, such as staff time, medical supplies, cost of travel for the patient, childcare costs for the patient, and costs falling on other social sectors such as domestic help from social services. Indirect costs are incurred by the reduced productivity of the patient, such as time off work, reduced work productivity, and time spent caring for the patient by relatives. Intangible costs are difficult to measure, such as pain of suffering on the part of the patient.

Recall bias pertains to how a person’s illness status can influence their tendency to report past exposure to a risk factor. Confounding arises when an additional variable is associated with both an independent and dependent variable. Observer bias refers to the possibility that researchers’ cognitive biases may unconsciously impact the results of a study. Publication bias refers to the tendency for studies with positive results to be more likely to be published. Selection bias occurs when certain individuals of groups are overrepresented, leading to inadequate randomization.

– The median is the point with half the values above and half below.
– In the given data set, there are an even number of values.
– The median value is halfway between the two middle values.
– The middle values are 56 and 58.
– Therefore, the median is (56 + 58) / 2.

Measures of Central Tendency

Measures of central tendency are used in descriptive statistics to summarize the middle of typical value of a data set. There are three common measures of central tendency: the mean, median, and mode.

The median is the middle value in a data set that has been arranged in numerical order. It is not affected by outliers and is used for ordinal data. The mode is the most frequent value in a data set and is used for categorical data. The mean is calculated by adding all the values in a data set and dividing by the number of values. It is sensitive to outliers and is used for interval and ratio data.

The appropriate measure of central tendency depends on the measurement scale of the data. For nominal and categorical data, the mode is used. For ordinal data, the median of mode is used. For interval data with a normal distribution, the mean is preferable, but the median of mode can also be used. For interval data with skewed distribution, the median is used. For ratio data, the mean is preferable, but the median of mode can also be used for skewed data.

In addition to measures of central tendency, the range is also used to describe the spread of a data set. It is calculated by subtracting the smallest value from the largest value.

In qualitative research, data saturation refers to the point where additional data collection becomes unnecessary as the responses obtained are repetitive and do not provide any new insights. This is when the researcher has heard the same information repeatedly and there is no need to continue recruiting participants. Understanding data saturation is crucial in qualitative research.

Stats: Correlation and Regression

Correlation and regression are related but not interchangeable terms. Correlation is used to test for association between variables, while regression is used to predict values of dependent variables from independent variables. Correlation can be linear, non-linear, of non-existent, and can be strong, moderate, of weak. The strength of a linear relationship is measured by the correlation coefficient, which can be positive of negative and ranges from very weak to very strong. However, the interpretation of a correlation coefficient depends on the context and purposes. Correlation can suggest association but cannot prove of disprove causation. Linear regression, on the other hand, can be used to predict how much one variable changes when a second variable is changed. Scatter graphs are used in correlation and regression analyses to visually determine if variables are associated and to detect outliers. When constructing a scatter graph, the dependent variable is typically placed on the vertical axis and the independent variable on the horizontal axis.