05.05.2023: If we use too few decimal places, the result will be imprecisely reported. On the other hand, if we use too many decimals, we give the impression that the result is more precise than it really is. How many decimals to use should be a conscious choice. Using more decimals than necessary is not wrong per se. However, an unnecessarily large number of decimals gives the impression that issues of measurement uncertainty or random variation have not been handled purposefully. Moreover, it conceals the message – the results are swamped by a surfeit of figures. The height of adults is normally...

09.12.2022: Not all data sets have explanatory variables and outcomes. The data may nevertheless contain associations that are worth revealing. In the 2010s, the Intervention Centre at Oslo University Hospital Rikshospitalet worked to develop a computer algorithm that could automatically detect tumours in a radiological image. The result of the computer algorithm was a two-dimensional geometric shape: the outline of a tumour. To check whether the algorithm worked, the outline generated by the automatic method was compared to outlines produced manually by four experienced radiologists. A geometric shape is...

24.10.2022: Attrition will occur in most clinical studies. If those who drop out differ substantially from those who participate, the generalisability of the study's results may suffer. Attrition can occur because not all of the invited participants choose to take part in the study, or because some of the included participants fail to attend the follow-up. Descriptive statistics In an attrition analysis, participants who have dropped out of the study are compared to those who were included in or completed it. For example, the Health Survey in the Department of Children and Youth, Division of Mental Health...

19.09.2022: Many methods for measuring levels of a substance in a sample have a lower detection limit. Data from these measurements must be handled in a way that avoids systematic errors. Let's start with an example: Figure 1 shows a fictional dataset of measurements of serum levels of a substance in two patient groups. The crosses indicate the actual values, but only those above the lower detection limit can be measured. We know how many values are below the detection limit, but not the actual values. These data are missing not at random (MNAR) because the probability depends on the non-observed values...

29.08.2022: Questionnaires often consist of scales composed of different questions and are intended to measure a given factor, for example anxiety or depression. Internal consistency is a measure of the strength of the relationship between questions in the same scale. When developing questionnaires and scales, we want them to be as reliable and valid as possible. In other words, the scales must be accurate, and they must measure what they are intended to measure. Internal consistency is a measure of whether questions that are intended to measure the same phenomenon are 'pulling in the same direction', and...

23.06.2022: When the outcome variable has only two levels or categories, standard binary logistic regression can be used. When it has three or more, we can use other variants of logistic regression. Let us start with an example: Munthe-Kaas et al. (1) studied a possible association between frailty before a stroke and degree of cognitive impairment three months after a stroke in 598 patients. The independent variable, frailty, was measured on an index that in theory ranges from 0 to 1, where a higher value indicates greater frailty. Before the stroke, the patients scored from 0 to 0.56, with a mean value...

19.05.2022: To compare proportions in two independent groups, one can use, for example, Pearson's chi-squared test. However, if the data originate from matched pairs, we must use methods that account for the dependence within the pairs. Matched pairs can originate from various situations and trial designs. The pairs may represent the status of a person before and after a treatment; a randomised controlled trial with a contralateral design where, for example, the left and right eye are randomised to eye drops A and B respectively; a matched case control study; or a crossover trial, as in the example below...

02.05.2022: Most randomised, controlled trials are conducted with parallel groups, but some treatments can be studied more effectively in a crossover trial. In a randomised trial with parallel groups, the participants are randomised either to treatment A or treatment B. In a crossover trial, each participant receives both treatment A and treatment B. Different treatment in two phases Let us start with an example where we studied the effect of a probiotic on irritable bowel syndrome in a double-blind, placebo-controlled trial (1). This was undertaken as a crossover trial, as illustrated in Figure 1. Such...

04.04.2022: The baseline value of an outcome variable is normally a strong predictor for the value of the outcome variable at a later time. In a longitudinal randomised, controlled trial, this should be taken into account. Let us look at an example: In 2012, all inhabitants of Trondheim aged 70 to 77 years were invited to participate in the 'Generation 100' exercise training study (1). Of 1 567 participants, 400 were randomised to high-intensity exercise training, 387 to moderate intensity and 780 to a control group that was recommended to follow the health authorities' advice on physical activity. The...

08.12.2021: Not all published research findings are reproducible — some because the findings are incorrect. What is the extent of the problem? A number of researchers have attempted to estimate how often published findings are false. They have used widely different approaches. Different methods The article 'Why most published research findings are false' by John Ioannidis attracted considerable attention when it was published in 2005 (1). The article was not based on data, but postulated a model for the proportion of false positive findings among published positive findings based on the following four...

22.11.2021: In randomised trials with measurements at two or more time points, it is recommended to adjust for the baseline value, for example by using an analysis of covariance. In an observational study, such an analysis will lead to biased effect estimates and problems with interpretation. When we randomise patients into two different treatment groups, the expected distribution of patient characteristics will be the same in both groups at the start of the trial. Any differences between the groups will be random, not inherent. In this situation, an analysis that adjusts for the baseline value will have...

28.09.2021: It is quite common to investigate multiple hypotheses in a single study, which increases the probability of Type I errors. This can be dealt with in various ways. A researcher may have various reasons for testing multiple hypotheses in the same study, for example to investigate the effect on several outcome variables, compare more than two groups or undertake separate analyses for sub-groups. Different adjustment methods Consider a study where six hypothesis tests are performed. If all tests are made at a significance level of 5 %, each of them will have a 5 % probability of making a Type I...