(2) basic principles of research design 2

Upload: jehad-ali-al-ammari

Post on 04-Apr-2018

220 views

Category:

Documents


0 download

TRANSCRIPT

  • 7/30/2019 (2) Basic Principles of Research Design 2

    1/12

    1

    Last time we said that bias is something that we dont want in our research, so

    any researcher tries to avoid bias, and we said that we have three different types

    of bias; selection, measurement and confounding bias. And all of these were

    discussed last time.

    Just quickly to define bias; its a process at any stage of inference tending to

    produce results that depart systematically from the true values.

    Slide (25): Chance

    Unbiased samples may misrepresent the population because of chance, so even

    if you could avoid bias, there is still something else to deal with, which is the

    chance, chance cannot be avoided like bias but we can minimize chance.

    Chance is the divergence of an observation on a sample from the true

    population value, itis called also random variation.

    The average height of Jordanian adult males is 170cm, but you may measure avery long guy with the length of 190cm, so this is chance, because the average is

    170 cm, and 190cm is far away from the average. So the presence of a very low

    percentage of tall people or short people is called chance.

    If you want to do any research, you have to minimize the possibility of chance

    by taking large samples; we dont want the chance to play a major role in our

    sample because this may lead to error.

    By tossing a coin 100 times, there is a fifty percent possibility to have the head,

    and fifty percent possibility to have the tail, but suppose that you want to toss

    the coin ten times, is it true that you all the time get five heads and five tails !!!

    Not necessarily, I may have seven heads and three tails, WHY? Because my

    sample was very small, but what will happen when I do it 1000 times? It will be

    very close to five hundred, so we can decrease the chance by increasing the

    sample size.

    E.G.

    E.G.

  • 7/30/2019 (2) Basic Principles of Research Design 2

    2/12

    2

    Slide (26): Chance vs. bias

    Bias distorts the situation in one direction or another

    In a research about the height of the Jordanian males:

    Bias: I only select tall people who are 185cm and over! Thats wrong (selecting

    bias), because I should choose the actual percentage of every category in my

    sample, so if I select all the people who are very tall, I will give the average of

    Jordanian male height about of 185cm which is not correct, this is bias.

    Chance: by selecting a random sample you have the chance of selecting taller

    people or shorter people.

    So if you distort or change your selection in one direction not both direction

    this is bias, in chance you have the possibility of selecting tall people or short

    people, thats why chance can be avoided by selecting random people and very big

    sample, you may get five percent of long people and five percent of short people

    and the rest will be around the average this is how we minimize chance.

    Chance or random variation results in an observation above the true value as

    likely as one below it. So you will have a chance of selecting people who are

    very tall, and you have the same chance of selecting people who are short, in our

    sample we all the time select very big sample to minimize the affect of chance

    - The mean of many unbiased observations of a sample approximates thetrue observation of the population, If your research is unbiased (by

    selecting random people without focusing on tall people) the outcome of

    the sample will approximate the true observation of the population.

    - In small samples this may not be close to the true observation of thepopulation, if you want to do a research the sample has to be large enoughto minimize the possibility of chance.

    CHANCE CAN BE MINIMIZED BUT CANNOT BE AVOIDED,

    BIAS CAN BE AVOIDED

    E.G.

  • 7/30/2019 (2) Basic Principles of Research Design 2

    3/12

    3

    Slide (27): Bias v. chance

    Bias can be prevented by proper conduction of clinical investigations, this

    means that you have to select properly by avoiding selecting bias.

    Bias can be corrected through proper data analysis, we said that we have

    different types of bias; selecting bias can be avoided before we start the study by

    proper selecting. Measurement bias and confounding bias can be corrected

    through the study or through the analysis of your study.

    Chance cannot be eliminated.

    Its influence can be reduced by proper research design.

    A student asked a question then the dr. said: thats a very nice question, if you

    took a sample, and there is a value which is far away from the average you

    cannot eliminate it, because this is chance, but we can minimize these values by

    choosing the sample randomly, because these values occupy a little percentage in

    our population.

    Bias can be deliberated (mota3ammed) for example I want to do a researchabout blood pressure among people who are smokers and I want to cheat, so I do

    bias, by this I get results that are not 100 percent true, so instead of choosing

    a random sample, I choose people who are low educated (usually smokers), bias

    can be deliberated to achieve the results that the researcher (cheater) wants,

    but chance is something present, so you have to minimize the effect of chance.

    Slide (28): Chance

    Statistics can be used to estimate the probability of chance or random variation;

    we can do some statistical tests to measure the percentage of chance. In general,

    the percentage of chance should not be more than five percent, some researchers

    say that it should not be over one percent, so in precise researches it shouldnt be

    over one percent, but usually most researches put a limit of five percent for the

    effect of chance on the results.

    Chance cant be eliminated, but its influence can be reduced by proper design of

    research, we've already discussed that.

  • 7/30/2019 (2) Basic Principles of Research Design 2

    4/12

    4

    Slide (29): Relationship between bias & chance

    This is a very nice paragram about the diastolic blood pressure, it is present in

    your BOOK !!! (WTH O_o !). It shows you the effect of bias. The average of

    the sistolic blood pressure is 120, and the diastolic is 80, this is if we measure it

    with the most accurate measurement, which is the intra-arterial canula. But if

    you decided to use another measurement which is less accurate like the

    sphygmomanometer, you will get nighnty, this is called messeurement bias, this

    bias has result in drifting from eighty to ninety. But in the case of Chance, you

    have the possibility of having messurements that are above and below theaverage, and the same percentage of above equal the same percentage of below,

    so by chance we have measurements above and measurements below, but by

    bias you move the data to one side.

    Slide (30): Validity

    validity means truth, we want the results to be true,

    Validity is correspondence to the true value measured or searched for, when I

    select my sample I want my outcomes or my results to be the true results for the

    population.

    I selected a sample of hundreds jordanian, to messure the hight of the

    jordanian males. I want the average of the jordanian male height in my results to

    be the same as the actual average if I selected all the six million people in

    jordan.

    E.G.

  • 7/30/2019 (2) Basic Principles of Research Design 2

    5/12

    5

    For an observation to be valid, it must be neither biased nor incorrect due to

    chance, now if you avoid bias and minimize the effect of chance, your results

    will be as true as possible, it will be very similar to the actual value of the total

    population.

    Slide(31): Internal validity

    It Is the degree to which the results of a study are correct for the patients being

    studied, you want your result to be correct for the sample you selected.

    Applies to the conditions of the particular group of patients being observed

    (your sample) and not to others (other samples).

    It is determined by how well the design, data collection and analyses are

    conducted and threatened by biases and random variation, what are the enemies

    of validity? Two enemies : bias and chance, all the time , when we want a

    research to be valid we have two enemies, bias can be avoided, chance can be

    minimized.

    Necessary but not sufficient by itself, is internal validity sufficient by itself ?

    NO, but its very necessarly, you want the actual values of your sample to be

    true, but this is not enough, we want the result also to be true for another sample,

    or another research on another sample.

    Slide (32): External validity (Generalizability)

    The ability to generize the outcome on the other samples that belong to the same

    populatoin

    It is the degree to which the results of an observation hold true in other settings

    It is the answer of:Assuming that the results are true in other settings, do they

    apply to my patients as well?

    Imagine that you are the person who did not do the research, and you are aware

    of the reserch, you want to ask your self, are the results of that research applied

    to my sample too? if the answer is yes, then there is a true external validity or

    generilazability for the research you want to take the results from.

    Generalizability assumes that patients in a study are similar to other patients.

  • 7/30/2019 (2) Basic Principles of Research Design 2

    6/12

    6

    A study with high internal validity may be misleading if its results are

    generalized to the wrong patients,

    If I do a research in the height of one hunderd males in jordan. And the result

    was an average of 170cm,can I apply that on jordanian females ! NO thats wrong.

    This is a very nice diagram :

    You have the possibility of selecting bias at the first stage, and then you have the

    possibility of measurment and confounding bias (for sure you have to avoid

    them), and you have to minimize the chance, finaly you have the conclusions.

    For the validity to be internal, the conclusions should be correct for the same

    sample. For the validity to be external, they should be correct to another sample

    belonging to the same population or to another group of patient not necessarly

    being examined, or to the total population.

    END

    E.G.

  • 7/30/2019 (2) Basic Principles of Research Design 2

    7/12

    7

    Designing a study (I)

    Research design means that you want now to do a research, you want to design

    your study.

    Slide (2): Choosing a research question

    The first thing to do is to choose a research question, which is the main idea of

    your research, the main problem of your research, it is not easy to select the

    research question

    First step of research design, first of all you have to identify a resarch question:

    - Determine the unknowns in your field, by reading your literature.- What do you wish you knew in your field?

    All the time as an anatomist I dreamed of a research in this field, which is now

    timing of eruption.

    What does the available literature lack? Your capacity and experience

    Research question

    Descriptive Analytic

    Slide(3): Descriptive research question

    Explaining clinical phenomena

    - prevalence of diseaseI want to meassure the prevalance of caries among school children, I want to

    describe it, go to the school children examine them and finaly I will say that

    the percentage is whatever, so this is dicriptive, I want to discribe

    something present, I am not going to creat something.

    E.G.

  • 7/30/2019 (2) Basic Principles of Research Design 2

    8/12

    8

    - Survival trendsI want to messure the prognosis for a man with prostate cancer who are

    alive at 5 years, my sample includes people who have done prostate

    surgery, or they have prostate cancer, and after five year I examine if

    they are alive, then I have to count or calculate the percentage of those

    people, this gives an idea about the survival age of that disease, this is

    also descriptive, I am not interfering, I am just discribing.

    - Health service utilization:I want to measure the percentage of seniors that are receiving H1N1

    vaccination, this is something about health service, people who are very

    old, dont like it to have vaccination, so for example I examine a sample ofone hundred old people, and then I have to ask him if he had H1N1

    vaccination and then I have to count, this is also discribtion I am not

    interfering; I an not doing an experiment, I am discription.

    - clinical testcharacteristicsI want to meassure mean value of LDL among patients with ischimic heart

    diseases IHD, this is also discriptive but it is done according to a test.

    but all of these examples they are discriptive, you are not interfering you

    are not doing an experiments, you are not giving that patient a drug to

    consume and then you measure the effect of that medication or drug , you

    just observe and describe.

    Slide(4) Analytic research question

    Is comparative:

    Is prevalence of caries higher among private or governmental school children?

    Now we have comparative, you dont only want to messure the prevalance of

    caries, you want to messure it among two different groups of people, children

    belonging to private school and children of govermantal school, you want to

    compare the results, this is analytical, because you want to compare.

    E.G.

    E.G.

    E.G.

  • 7/30/2019 (2) Basic Principles of Research Design 2

    9/12

    9

    More significant than descriptive questions

    because by its answer you get the conclusion of the research, and the answer

    enable us to develop intervention to prevent disease, and Target intervention to

    particular population

    if I examine the caries level among private school and govermental school

    children, and I found that the ammount of caries in the govermental children is

    higher, this mean that something is wrong, so I will try to find the differences

    between private school and govermental school, it could be that the govermental

    school children have lower rates of health, so I have to increse the health care

    in these group of children to decrease caries, but maybe caries in private schoolis higher, because in private school children become higher poket money so more

    chips and chocolates, what should I do ? I have to interfere with that !

    Descriptive questions must be answered first

    All the time we have to answer the discreptive question first, then the analytic

    question.

    I have to do a research for caries among govermental school alone so this is

    discriptive, and I have to study the ammount of caries in private school alone

    too, and then I have to study both types and compare between them,

    About the research I am doing now, timing of tooth eruption among jordanian

    school children, after I finish this research, I can do a reseach about timing of

    eruption for those who have certain diseases like cleft lip or cleft palate, then I

    compare the timing of eruption of those people, and compare it with the timing

    of eruption with normal people, this is analytic, I can't do that before I do the

    first research for the normal people.

    E.G.

  • 7/30/2019 (2) Basic Principles of Research Design 2

    10/12

    10

    Slide(5):Choosing a research questions

    Specify the population

    Determine the length of the study

    if I knew that I will stay in jordan just two years, I cannot do a research that

    takes five years !!!

    Or a master student will make a research, and the research needs two years,

    and he have to do it in six month !

    Your willingness towards completion do you really want to do this research to

    the end, or just just to get promoted from assistant professeor to associate

    professor !

    J.P. Ioanndis did a research and found that the Median time from enrolling

    subjects (start to inviting people to participate) untill the publication was found

    to be 5.5 years (five years and six month is the average of researchs )

    Slide(6):Choosing a research questions (1st criterion)

    Choose a question that keeps you excited all the way through; HOW ?

    - Identify obstacles that are preventing to do the research, in order toperforme and publish the research.

    after finishing a research no one accept to publish it ! this is an obstacles !

    then I get frustrated and I will not be able to do another reserch :S :S

    Examples of obstacles:

    - Subjects: noncooperative patients, or patients who agreed at the first timeto participate, and then refused !

    - ethical approval: sometimes you may not get the approval (mowafaqa25laqeyye), this approval is important to protect the rights of the people.

    you cannot ask a big ammount of healthy people to get examined under a

    chest x-ray, they will get exposed to radiation !!!

    so the research have to be approved first by research commities

    E.G.

    E.G.

  • 7/30/2019 (2) Basic Principles of Research Design 2

    11/12

    11

    - collaborators: people who assists you, I cannot select a person to assistme and then I discover that this person is not a good person, or he just

    broke up with the research

    - lab capacity: are the instruments and devices available in the lab ?I asked the JUST university to provide me with a special 3D camera to

    continue my researches, but it was not available in JUST.

    - review committees: people who review my publication.- Editors.- missing data: do I have the ability to save my data from being lost?

    Slide(7): Choosing a research question (2nd criterion)

    Choose a question that will have an impact on the health and well being of thepopulation: your research has to be significant

    - Difficult to fully appreciate the impact a study will have before doing it,andsometimes during your research, you become new ideas, which will

    effect on your conclusions, for better results.

    - Purpose of clinical research is to improve health not for grantsmanship,publication and promotion

    - Enrollment of sufficient number of subjects, all the time think about anenough sample size.

    Slide(8): Choosing a research questions (3rd criterion)

    Consider what questions you are ready to answer based on:

    - The prevalence of the disease in your area: like diabetes in jordan.- Your prior experience- Your colleagues- Your community contacts

  • 7/30/2019 (2) Basic Principles of Research Design 2

    12/12

    12

    Slide(9)Choosing a research questions (4th criterion)

    Be sure it has not already been answered unless if you can do better, or you

    found a deffect.

    - Computerized literature searches: to know that a research has not beendone yet.

    - Consulting the others in the field: you should ask your colleges if theresearch has been done yet, if yes, dont do it again, there is no need.

    - Attending conferences: and ask people there. Scientific and academic relations Not all conference abstracts are electronically accessible: so

    not everything is vissible in the internet; so you have toattend conferences to know if the research that you are going

    to do hasnt been done yet.

    THE END

    The last lecture was sopposed to be 15 pages, I was surprised when I saw the shift in the

    pages :S :S . the normal one is available on my toothy.

    I dont know about you, but I like this simester

    DONE BY:

    AMMAR ANAGREH