Upgrade to remove ads
Only $35.99/year

Notes from revision class research methods

Terms in this set (14)

Experimental trials?
RCT and controlled trials
What are the strenghts of RCT?
Strict, extra level in it in relation to the randomisation which aims to reduce the bias. Sometimes not always relevant to randomise them (medically)
What are controlled trails?
they are experimental

They aren't randomised, selected and put into groups which means bias could come into this stage
What are the observational studies?
Cohort, cross sectional, case study / case serie
What is a single case study?
Looking at one case only

Strengths - lots of in depth information about that single case, lots and lots of depth

Weaknesses = reduced generalisability because only looking at one case
What is a case series?
looking at a number of case studies, starting to group together single cases to see if there is any patterns between the numbers of people rather than just an individual person. Picking 10 people, thinking about them individually first (case study) but then look together at all 10 to see patterns. This increases in generalisability, as start to present data about a large amount of people.
Cross sectional study ?
Polling, snapshot at one moment in time. questionnaire, survery at one moment in time. Elections = a cross sectional study.

Strengths = gives an idea of currency at that moment in time, what is happening at that moment in time, flavour and snapshot. Might be really important as might want to know at that moment.

Weaknesses = its not a whole picture, only stands for at that moment in time. Might get a different picture even a couple days a part. Poll's might not reflect the final outcome.
Cohort studies?
more longitudinal type of aspect to it, look for patterns over time. Look for associating and patterns overtime. Good for flag raisers, can raise flags that we can then go and investigate more specifically in an experimental study. Patterns and trends coming through from a group of people.

Limitation = can't control the variables in an observational study, can only sit back and observe. All sort of things could influence the results.

Can get lots of data from observational - especially if its longitudinal.
Experimental - unrelated different group design
2 different groups of participants, that we hope are matched as possible, who are doing the interventions. Is there a difference between the 2 interventions? Because we have 2 different groups of people doing it, need them to be as matched as possible. This is important because we need to make sure that it is just the independent variable that is going to shape the results at the end. Because they are different groups need to make sure that there is nothing that can influence it
Experimental - same group design =
= one group doing one intervention at one time. Don't have to worry about matching as it's the same group of people so know other characteristics that are going to influence the results. However the order of interventions may influence / impact the results, as you're doing one before the other. To try and reduce the order effect - split the groups in half, 50 does one first, 50 does the other first. And then they swap. Could have a break in the middle - washout period - to take away any of the influence of the first intervention so it doesn't influence the second intervention. Appropriate on some designs but not all as to whether you can have a washout period in the middle.
When to do a parametric test?
Parametric (looking for the parameters of our data) does the data we have fill certain parameters or not - if yes do a parametric test. If no do a non-parametric test
2 fundamental = normal distribution, interval or ratio level data. If they have these parameters can do a parametric test. If you don't, do a non-parametric test.
Parametric test = more robust, more rigorous, more difficult to get a statistically significant result. Confidence in this result is increased if get a statistically significant result - because it was harder to achieve, not due to chance.
Non-parametric test = less robust, easier to get a level of significant, less confidence in our findings in our results.
What is the P-value?
we set our p value at at 0.05 because it gives us a 1/20 chance that our results have occurred by chance. Give ourselves a little bit of leeway, happy to be 95% confident because were getting data only from a sample of people. Can never be 100% confident that it 100% represents a true population - unless we get data from the entire population. Gives ourselves a small amount of leeway so that we have confidence that were aware that the sample may not truly match the population, but 95% is close enough. In a different sample get different results maybe.
If we have a level of statistical significance we still need to know whether it is clinically significant. Further tests need to be done to know if it is clinically significant. Something can be statistically significant but completely irrelevant in terms of clinically significant.
Outcome measured 'research gathered' strengths and weaknesses
berg balance = strength, chance of being more objective - particularly if you're blinded a blinder assessor or if you've got a tight protocol checklist, intra-reliability as long as standardised, consistent results, reduces researcher influence bias. AS long as you've enhanced the robustness
Weaknesses = patient may still bring in bias, might chose an outcome measure that will skew results more in your favour, It might be completely meaningless to the patient. Have to bear in mind this. Balance = physiological and psychological component, fear of falling etc. berg balance might not indicate psychological factors as to why they're falling.
outcome measured 'patient reported' strengths and weaknesses
strengths = more meaningful for patient, hearing the patients voice, its their report, their perceptions,
Weaknesses = much more subjective, nothing objective about this at all. Because it subjective it might be influenced by mood, or about the participant when they was filling it in, hunger, temp. , attitude etc.


In research it is nice to do a bit of both. Potentially allows you to look at them for patterns - for example berg balance and their confidence.

Also it is dependent on what your research question is, is it about patients confidence in balance - patient led measures are going to have greater strengths from that time.
Other sets by this creator