No. The service is free for all eligible customers At least for now, for anyone who is not eligible, the service is not available: there is currently no possibility for us to serve external clients, even for a charge.
The service is available for members of Helsinki University’s Medical Faculty, HUS, or FIMM only. Unfortunately, we cannot currently take on clients who do not have an affiliation to one of these departments/institutes.
It depends on our workload, which can vary substantially. Generally within a few working days, but it may take up to a week. Booking in a meeting after that can of course take longer, depending on the consultant’s and client’s relative availabilities.
This also depends on our variable workload, and of course on the complexity of the question. In general, the time it takes to arrive at a full solution to a customer’s issues ranges from 30 mins to a few hours. However, we have plenty of cases where the question takes more time, or the customer needs regular support over a longer period, or the consultant becomes more deeply involved with the research project, all of which can lead to a case taking much more time. Currently we have no hard limits on this aspect of our service.
Usually, we work using a combination of meetings and email exchanges. Meetings can be remote or in person at our office in Biomedicum 2. The balance between meetings and emails varies a lot and depends on the customer’s preference, time budget, everyone’s availabilities, and the complexity and type of questions.
No. We are an advisory service, which does not extend to actually conducting the analysis. In exceptional cases, the consultant may become so involved with a particular research project that they become in practice a member of the research group and correspondingly feature as a co-author on the paper, but such cases are rare.
No. For data security and privacy reasons we do not receive data. Instead, bring your computer with the data on to the meeting, so that consultant and client can look at it together then.
Yes. We provide advice on any aspect of your study that concerns the handling, analysis and interpretation of data.
It depends. If it is a short snippet that is easy to understand outside the broader script, then yes. If it is a whole script that needs “proof-reading” or debugging, then no. Where the boundary goes between these two is hard to say and varies from case to case.
Yes. We can help you with phrasing, clarity of language, writing a few sentences and proof-reading to make sure the text correctly represents the statistics/analyses that were done. However, we will not write the whole section for you. The final responsibility for correctness always remains with the researcher, not the consultant.
Absolutely. We only ask that you have done some investigation beforehand about what techniques might be suitable for your questions and data, so that we have somewhere to start from.
Analysing, maybe, depending on the techniques you want to use. Processing, unfortunately not - this is a task for specialist bioinformaticians.
Check out our useful links too! They contain many informative resources.
We can’t directly teach you statistics or relevant software, because the time we have for consultations is limited. We can, of course, provide pointers and guidance in your own learning as it pertains to your specific research needs. Get started by coming to one of our courses, using a classic textbook, or check out our useful links page for a wide selection of learning resources. For help getting started with coding in R, try our kickstart guide.
This is a question that indicates that you are not very familiar with basic statistical techniques and thinking. In order for us to help you, we ask that you first learn some of these basics - see above for good places to start. Once you understand the basic concepts, if your data and questions are not too complex, you can use clear criteria to identify the appropriate test. If you still are unsure, get in touch with us.
As above, this is a question that indicates that you are not very familiar with basic statistical techniques and thinking. Calculating a p-value requires a specific statistical test or model, which you need to decide on first, and it requires variation in the data (i.e. if you only have 2 unique values, you cannot calculate a p-value for how different they are). In order for us to help you, we ask that you first learn some basics of statistics - see above for good places to start.
A correlation quantifies how closely two variables are associated with each other. A regression additionally quantifies how steep that relationship is. So only with a regression can you say e.g. “For every year of patient age, blood pressure increased by X mg Hg on average”. A correlation would only say “Age explained 50% of the variation in blood pressure”.
Generally, no. For a t-test, the distribution of the response variable should be normal within both study groups. For an ANOVA or regression, what needs to be normal is the residuals of the model, not the raw data - so you might have data that is not normally distributed but once you fit a suitable explanatory factor, the residuals are still OK. If your residuals in these analysis are not normally distributed, you cannot use that model.
Unfortunately not. Clustered or non-independent data (e.g. a dataset that includes several families, each represented by several patients, or a design where each patient has gene expression in different tissues quantified) often includes variation at the level of the cluster - e.g. the outcome variable of one family member is more likely to be similar to other members of the same family than to unrelated patients - which has to be accounted for. If not, what often happens is called pseudoreplication: you think you have 60 data points, because you have 60 tissue samples, but they were from only 12 patients. Using n = 60 would artificially make your confidence intervals smaller, which risks giving you spuriously significant results. In actual fact, your data only lets you understand variation between 12 patients, not 60 tissue samples.
These things might also help:
Googling your question. Try different search terms, being more (or less) specific, including the name of both the relevant test and software, putting search terms in a different order, etc.
Our useful links page. The resources there have more detailed information on a broad range of topics
Just try some things in your software, for fun - graphs, descriptive statistics, simple tests etc. This will probably help illustrate what is going on in your data, what analyses are appropriate, and where problems might arise.
If it feels like you've tried everything but still are not making progress on your own, fill in our e-form to book a personal consultation (for in-depth questions: c. 1-hour meeting slots) or come to a drop-in workshop (for shorter questions, c. 15 mins per customer), held every other Wednesday.