Parametric vs Non-Parametric Statistics

Statistics are used to draw conclusions about a population based on a sample. A parameter is a numerical value that describes a population characteristic, but it is often impossible to know the true value of a parameter without collecting data from every individual in the population. Instead, we take a sample and use statistics to estimate the parameters.

Parametric statistical procedures assume that the population distribution is normal and that the parameters (such as means and standard deviations) are known. Examples of parametric tests include the t-test, ANOVA, and Pearson coefficient of correlation.

Non-parametric statistical procedures make few of no assumptions about the population distribution of parameters. Examples of non-parametric tests include the Mann-Whitney Test, Wilcoxon Signed-Rank Test, Kruskal-Wallis Test, and Fisher Exact Probability test.

Overall, the choice between parametric and non-parametric tests depends on the nature of the data and the research question being asked.

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 negative SMR is not possible. An SMR less than 1.0 suggests that there were fewer deaths than expected in the study population, while an SMR of 1.0 indicates that the observed and expected deaths were equal. An SMR greater than 1.0 indicates that there were excess deaths in the study population.

Calculation of Standardised Mortality Ratio (SMR)

To calculate the SMR, age and sex-specific death rates in the standard population are obtained. An estimate for the number of people in each category for both the standard and study populations is needed. The number of expected deaths in each age-sex group of the study population is calculated by multiplying the age-sex-specific rates in the standard population by the number of people in each category of the study population. The sum of all age- and sex-specific expected deaths gives the expected number of deaths for the whole study population. The observed number of deaths is then divided by the expected number of deaths to obtain the SMR.

The SMR can be standardised using the direct of indirect method. The direct method is used when the age-sex-specific rates for the study population and the age-sex-structure of the standard population are known. The indirect method is used when the age-specific rates for the study population are unknown of not available. This method uses the observed number of deaths in the study population and compares it to the number of deaths that would be expected if the age distribution was the same as that of the standard population.

The SMR can be interpreted as follows: an SMR less than 1.0 indicates fewer than expected deaths in the study population, an SMR of 1.0 indicates the number of observed deaths equals the number of expected deaths in the study population, and an SMR greater than 1.0 indicates more than expected deaths in the study population (excess deaths). It is sometimes expressed after multiplying by 100.

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.

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.

Validity in statistics refers to how accurately something measures what it claims to measure. There are two main types of validity: internal and external. Internal validity refers to the confidence we have in the cause and effect relationship in a study, while external validity refers to the degree to which the conclusions of a study can be applied to other people, places, and times. There are various threats to both internal and external validity, such as sampling, measurement instrument obtrusiveness, and reactive effects of setting. Additionally, there are several subtypes of validity, including face validity, content validity, criterion validity, and construct validity. Each subtype has its own specific focus and methods for testing validity.

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 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).

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 relative risk (RR) of the event of interest in the exposed group compared to the unexposed group is 2.5.

RR = EER / CER
EER = 10 / 200 = 0.05
CER = 10 / 500 = 0.02
RR = EER / CER
= 0.05 / 0.02 = 2.5

This means that the exposed group has a 2.5 times higher risk of experiencing the event compared to the unexposed group.

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.