Designing of research study should be given more attention than on data collection and analysis. It is difficult to re-do the poorly designed study but you can always re-analyse data with different method. Therefore, it cannot be stressed enough that best time to contact ISCON statistician is before designing your research study.

The studies with poor planning i.e. vague research questions, unclear hypothesis, inadequate sample size, inappropriate choice of variables, improper randomisation, poor follow up are deemed to produce inaccurate and meaningless results. Statistician at ISCON Statistics can be your best companion help every steps of your research project – starting from research idea to development of study protocol to successful completion and reporting in high-impact journals. Our statistician can help you with-

  1. Formulating research questions and research hypothesis and develop statistical analysis plan
  2. Defining and clarifying the study outcomes- primary outcome, secondary outcome, data collection methods and frequency of the measurements
  3. Advice you about study design whether it’s appropriate to answer research questions or hypothesis
  4. Discuss with you the impact of bias, randomisation, allocation and blinding on results
  5. Discuss with you the required sample size (so study has optimum statistical power) in different conditions/hypothesis
  6. Help you to develop study protocol, statistical analysis plan, any secondary or interim analysis.

Help translating your research idea into a statistically answerable question

In planning research study, it very important to formulate research question that is clear, specific and well defined. Because, it is your question that will determine the conduct of the study and method of statistical analysis. It is your research question that will guide the development of the study protocol, sample size estimation and power calculation. Well defined research questions have three essential characteristics. First, the research question should describe or hypothesize the relationship between two or more variables. Second, it should take the form of question and third, the variables in proposed questions must allowed to be observed or measured through either observation or by experiments. A vague research question not only leads to complexity of both data collection and subsequent statistical analyses but also indirectly affects reliability of the proposed study. The specific and clear research questions help statistician determining the sample size, power and method of the analysis - producing valid and meaningful results of your study. ISCON Statistics will. ISCON recommends PICO format to develop specific research question: Population of interest, the Intervention being studied, the Comparison group (or to what is the intervention being compared) and the Outcome of interest.

Let us help you to derive clear research question from your research topic so that your research project conduct smoothly- drawing meaningful and accurate conclusion.

Help in articulating specific research hypothesis that drives your statistical analysis

Your research question helps you to articulate specific research hypotheses. Hypothesis is basically a statement – a educated guess about potential relationship between two or more variables. Formulating a clear research hypothesis based on specific research question drives the smooth conduct of the study ultimately leads to valid data analysis. Only after research hypothesis being formulated, data will be collected and then tested statistically whether to reject or not to reject hypothesis, based on observed sample data. The research question and hypothesis need to be formulated before the start of the study, not after data being collected. It a research questions that should determine the hypothesis, not the observed data. Retrospectively formulating hypothesis from observed data carries risk of multiple statistical testing which than leads to declaring dubious statistical association.

In statistics, hypothesis can be null and alternative hypothesis. The null hypothesis states that there is no association, difference or correlation between two or more variable whereas alternative hypothesis states that there is association, difference or correlation. The statistical test or model than test this hypothesis whether to reject of the hypothesis. Hypothesis can be directional (one sided) or non-directional (two sided). In directional hypothesis, one has to specify not only the association but also direction of association-difference or correlation (how much greater, lesser or better or worse) in hypothesis statement.

From your research idea, we help you to set clear research questions and formulate testable research hypothesis.

Helping you choosing right variables for your research

Variable is a characteristic that varies from person by person or item by item, i.e. takes different value in different person, item or thing. There are thousands of variables can be measured and reported in research literature, the question is which variables to choose and which to discard, if chosen too many, how it would be included in statistical analysis.

Variables can be dependent or independent. The dependent variable is the variable which values are dependent on other variables- which are called independent variables. We are interested in assessing the effects of the independent variables on dependent variable (outcome). The better nomenclature would be - response variable and explanatory variables. We are interested is how explanatory variables (independent variables) explains the variation in response variable i.e. outcome (dependent variable).

The choice and number of variables in your study depends on multiple factors. It depends on your study objective- whether you are interested in specific phenomenon (limited numbers) or identifying important variables from collecting large number of variables (through specialized statistical methods). It also dependent on your subject specific knowledge – include variables that are important in a way that are associated with outcome and/or explanatory variables or both (known as confounding variable). Omitting or not handling confounding variables appropriately in statistical analysis often leads to biased and inaccurate conclusion. ISCON Statistician can help you to choose optimum number of variables and analyse your study with best possible methods getting you valid and unbiased statistical results.

Help in calculating sample size required in your research study so that your study has optimum statistical power

Sample size is number of participants needed in the study to answer research questions. The aim of sample size determination is to calculate number of samples required in study in order to detect important, meaningful effect. We needed to calculate this number with great precision because if the sample size is too small, we many not be able to detect an important effect, whereas unnecessarily large sample size leads to waste of resources. So, it is so vital to optimize the sample size during planning phase as you want to make sure that effort you are making in research study is worthwhile i.e. should leads to valid and meaningful conclusion. It is also ethically undesirable to expose more participants than needed from side effect, toxicity of drug or intervention. Therefore, majority of ethics committee always ask about the optimum sample size in your research study before they give you approval.

In order to determine correct the sample size, you need to have clear research question, set of clear study hypothesis, details about your outcome variable, its variability and expected/predicted change in your outcome in your study. Specifically, you need to have details about four main items (1) the extent at which you want to have false positive results in your study (alpha which type 1 error rate, most commonly 5%) (2) probability that you want to detect the effect when it really exist (statistical power, commonly 80%) (3) smallest difference that clinically meaningful i.e. effect that you don’t want to miss, also known as minimal clinically significant difference(MCSD) and last (4) variability of outcome in population in which you going get sample (which you can gather from previously published studies).

The formula and method of sample size determination dependent many factors such as - study objectives, number of hypothesis, type of outcome (continuous, binary, time to event, count), number of group or explanatory variables in your study, type of your study design (observational or experimental study, parallel group clinical trial, cross-over trial, cluster or adaptive trials), method of statistical analysis (simple statistical test or advanced statistical models with covariates included such as liner mixed model with clustering) and resources (funding source, recruitment rate, staff availability, timelines).

Its is prudent to ask statistician to calculate optimum sample size because the complexity involves in methods and scenarios and secondly, statistician may have expert knowledge of other efficient, specialized method (such as Monte Carlo simulation) than formulas which can establish optimum number of samples robustly. ISCON Statisticians particularly experienced in calculating and explaining the sample size calculation in different scenarios in wide variety of study designs and statistical methods. Get in touch with us for more details.