Homework for my TA, week 4

These are the links to the blogs i have commented on this week:

http://liamjw91.wordpress.com/2011/10/12/%e2%80%9cis-it-possible-to-prove-a-research-hypothesis%e2%80%9d/#comment-22

http://psud5d.wordpress.com/2011/10/12/stats-blog-3-wildcard-is-google-killing-our-brains/#comment-10

http://standarderrorofskewness.wordpress.com/2011/10/13/why-is-reliability-so-important/#comment-13

http://itsafreudianslip.wordpress.com/2011/10/14/week-3-of-the-weekly-blogs-how-do-politics-affect-the-science-of-psychology/#comment-18

Reliability and validity and their influence on experimental control

Lack of control in an experiment means findings are invalid, even if they are reliable. Therefore they can’t be generalised to the wider population. This is problematic when we consider generalising the effectiveness of a drug. However sometimes you need to take a gamble and hope there aren’t any influential extraneous variables as too much control would mean science wouldn’t ever progress!

Reliability refers to how consistent your findings are (Dictionary.com, 2011) and validity ensures research is actually measuring what it set out to measure (allpsych.com). You can have a reliable study with invalid findings but not a valid study with unreliable findings. If you’re testing what you think you are testing then you will get a consistent result. This shows validity and reliability. But you could get a consistently wrong result. This finding would show reliability but not validity.

For example, your hypothesis is that all students prefer fish and chips to strawberries. You interview 70 students before lunch but have to ask the other 30 after lunch because you run out of time. Your findings show students prefer fish and chips to strawberries. This result is reliable because you find a consistent result. However it is not valid because extraneous variables (hunger) have influenced the outcome rather than the manipulation of independent variable (strawberries or fish and chips.)

This experiment would have to be redesigned because the more extraneous variables there are, the less control there is in the experiment. This means the less validity the study will have. This becomes a problem when needing to generalise findings to the greater target population. If extraneous variables have made your findings invalid you may, for example conclude a drug reduces depression- when actually it doesn’t.

This is a serious problem, however controlling all potential extraneous variables is impossible. Some things cannot be predicted- for example a lecturer couldn’t have predicted seeing a photo of an eye could make someone feel sick. If you have too much control we would never do anything! So sometimes you need to just take the risk and believe it is your independent variable causing the results and not an unexplained variable. Taking the risk and getting it finding extraneous variables is better than not taking the risk at all… if no risks were ever taken, no experiments would be conducted and science would never progress!

 

Dictionary.com. (2011). Retrieved from http://dictionary.reference.com/browse/reliability

Allpsych.com. (2011). Retrieved from http://allpsych.com/dictionary/v.html

Homework for my TA, Week 3

These are the comments i have made and would like marking please:

http://afshinpsychology.wordpress.com/2011/10/06/do-you-need-statistics-to-understand-your-data/

http://liamjw91.wordpress.com/2011/10/07/do-you-need-statistics-to-your-understand-data/

http://emcg1.wordpress.com/2011/10/07/%e2%80%9cdo-you-need-statistics-to-understand-your-data%e2%80%9d/

http://statisticalperrin.wordpress.com/2011/10/07/do-you-need-statistics-to-understand-your-data/

Thank you!

What makes a research finding important?

All research is important. Every finding is important to someone, but not everyone. However the research funded by the government often gets the most media interest, because they choose topics that appeal to the mass audience therefore appear to be more ‘important’ than other research. But just because a finding isn’t widely acknowledged does not mean it is unimportant.

When Einstein discovered gravity he thought his findings were important. When Fleming discovered penicillin, he thought his findings were important. If a researcher of tomorrow discovers the cure for Alzheimer’s, he will think his findings are also important. Every researcher perceives their own work as important and credit worthy and wants to be rewarded for their findings. Often in society however, some research is dismissed because it doesn’t fit with current beliefs. For example evidence suggesting the world was round was disregarded until enough research had been presented so it became impossible to ignoring the findings. That first researcher will have believed his work was important and yet officials cast it aside. It is a safer choice to propose research that matches common thinking than opposes it, because it is more likely to receive funding. Findings that are similar show reliability, so act as ‘safer’ investments for government bodies who decide how to distribute their pot of money.

The government also have the general public to satisfy so need to invest in research that will provide the greatest results for the greatest number of people. Therefore the cure for Alzheimer’s is deemed more important than a deadly disease attacking elephants for example. I am not saying Alzheimer’s is less important than other causes, but it is only more important to some people. To the researchers investigating elephant illnesses, their findings will be equally as important.

Everything is important to someone, but not everything is important to everyone. Therefore with research, often those presenting the least controversial ideas and offering the greatest benefits to the majority of people are perceived as most important, because there is only a limited amount of money to be spent on research investigation and priorities have to be made. It doesn’t seem fair…

Do you need statistics to understand your data?

Not all data can be analysed using the same method of statistics, but all data can be analysed. Data that is left unanalysed is useless; Rows of numbers, or categorises of names that do not show us anything. I understand that some people may argue simply grouping data or implying correlations is not classed as statistics because it is not difficult maths to do. But if it is not statistics, what is it?  Any action taken upon a sample of results in order to enhance understanding is surely a type of statistical analysis, no matter how easy or difficult. From grouping all categorises that have an ‘S’ in their name, to finding P values, F values and cohort effects- it is all statistics! Anything done to data is statistical because it changes our understanding of what the data means.

To show my point more clearly I am going to define the key words within the title to show how statistics are needed to understand data. The definition of the noun ‘need’ is to have a requirement, duty or obligation to do something (dictionary.com). ‘Statistics’ means having the ability to collect, analyse and interpret data (Oxford English Dictionary), and ‘understand’ means to perceive the full meaning of something (dictionary.com). So if we were to reword the initial question incorporating these definitions, this new question could be: Do you have an obligation to collect, analyse and interpret information to perceive the full meaning of your data? For example: do scientists have an obligation to collect information and analyse it, in order to find trends that predict cures for cancer? Yes. Do weather forecasters have an obligation to collect information and interpret it, in order to predict future weather patterns, so hurricane evacuation plans can be executed? Yes.

Statistics are used to predict results and find trends within data. Any data that goes unanalysed could have been that key bit of information needed to help find medical cures or warn people about future problems that could arise. Statistical Analysis could save lives!