conclusion questions confuse me...
like for a question like "what can we concluded from the results of this study" idk what to say!... some answers say none can be concluded because it is a descriptive statistics, some write it can be concluded that there is a big difference between the two groups and some write none because it isnt statistically significant..
so im completely confused...
descriptive stats cannot infer conclusions. for a conclusion to be made, inferential stats must be used, and these must be statistically significant.
So conclusions can only be made if there are p-values, where the p-value is less than 0.05 otherwise no conclusions can be made, right?
yeah, but although p-values are the only inferential statistics you need to know, they aren't the only type.
Well, strictly speaking, p-values aren't a 'inferential statistic', but a product of one. The inferential statistics is things like the t-tests, ANOVAs and the Chi-Squares. All these tests produce p-values of some description.
you're creating an operational hypothesis for a research study that has already been completed and you have all details of the experiment and its results in front of you. your operational hypothesis should reflect the results as usually the results in those questions reflect what the researcher was expecting
e.g. if they were testing alcohol consumption and driving ability, and drivers with lower scores were those who drank alcohol, it makes sense to have your operational hypothesis utilize those results and state "18 year old males who don't drink alcohol before undergoing a test in a driving simulator will perform fewer errors than 18 year old males who have 3 standard drinks before undergoing a test in a driving simulator
you utilize the results without stating what the results were. make it so your operational hypothesis agrees with the results, and i'm pretty sure you won't go wrong.
NO NO NO NO NO!
Your hypothesis MUST have been based on the research that you have done, NOT the results of your experiment. The reason for this got to the heart of why you're experimenting in the first place. What you are doing is looking at previous results, designing the experiment so that it fits in a different context and seeing whether you would get the same results. You can't do that if you're using your own results as your hypothesis. Also, it's the contradictions between your hypothesis (and basically what the world thinks would happen in your experiment) and the actual results that actually makes the science.
yes, I agree with you here and in my E.R.A sacs, we always created the hypothesis before conducting the experiment (let alone interpreting the results)
That's the proper way of doing it. Otherwise it's possible to use data dredging techniques to come up with hypothesises that aren't actually supported.
ah okk....
then if a question asks you, is the hypothesis supported?? then can you say yes or no just based on the descriptive stats as this isnt making a conclusion?
Generally speaking, you can't really make any statements about how much support a hypothesis has just on the basis of the descriptive statistics alone - in fact it's pretty poor science to do so. The reason for this is simply. Take for example I do an experiment on whether tiredness reduced the ability to recall words remembered. I use a Karolinska Sleep Scale (KSS) (This is a test used to measure sleepiness in Psychology) and that I separate into a sleepy group and the awake group. Suppose then I do the recall bit three time and I get the following results (each row of trial is the average of that trail):
| Sleepy | Awake |
1 | 5 | 7 |
2 | 8 | 4 |
3 | 2 | 7 |
Mean | 5 | 6 |
Using descriptive statistics, you might say that make a conclusion that being awake means you're more likely to be awake than when you're asleep. But look at the difference, it's only by a value of one. So you have to ask yourself, how sure can you be that the one is actually significant - that the difference in values actually support the hypothesis that you're looking at, rather than being a product of something else (such as chance). [The strict definition of a p-value (which is the measure of significance used) is the probability of getting a result that is equal or higher than the result that you got , assuming the null hypothesis (which is the opposite of your hypothesis) is true]. This is why you need to use inferential statistics to make determination os wheither the results support your hypothesis.
A conclusion is a statement about whether the hypothesis has been supported or not right??
how can inferential stats like t-test/p-value determine whether the hypothesis has been supported or not?? i mean if you look at graphs/tables (descriptive statistics) then it would be pretty clear if the IV has made a difference..but of course the p-value shud be below .05..
You're correct on the first question.
To the second question, this is actually very hard to explain without going beyond the VCE course and into first-year university territory. I'm going to have a crack in it anyway (I've always had a disliking to teachers that say that something is 'beyond the VCE course'). Depending on the amount of mathematical background that you have (you'd probably need to have done or are doing Mathematical Methods 3&4 or Further Mathematics 3&4 to not get lost), you'll either completely follow what I'm writing here, or end up in bat-shit confused territory. Anyway, here it goes.
Psychology uses a lot of statistical tests in research. Many of these are classed in a set called 'test of significance' and it's often used to determine whether the differences between means is something worth writing home about (and writing journal articles about it), or just dumping the research onto the cabinet. These tests include things like the Student's t-test, Analysis of Variance (ANOVA), Multivariate Analysis of Variance(MANOVA), and other statistical tests with increasingly larger acronyms. These get more and more complicated, but the general principles of how they work is the same.
Now, here's the maths bit of it. I want you to think of a bell curve. If you can't think of one, here's a picture of one to help.
From: http://classes.kumc.edu/sah/resources/sensory_processing/images/bell_curve.gifThose of you with the maths background should know that this bell curve represents data, much like a graph that has been collated together. Now think of this curve as representing all of results that the participants got for a particular condition (e.g. the experimental condition) that your participants went through. If you're struggling here, imagine there's an imaginary line connecting the curve to the horizontal axis. This imaginary line represents the result one participant got. Now imagine that there are multiple lines doing the same thing. Eventually, given a large enough sample (in mathematical terms.
, but in practical terms, a large enough sample), that should become that curve you see there. Now, imagine another bell curve. This represents all the results that your participants got for another condition (e.g. the control). Basically, when you dump the curves onto the same horizontal axis, you're going to get overlap between the curves (you're going to get participants from one condition who will get the result as participants from other conditions). What 'test of significance' measure in a sense, is the degree of overlap between curves. This is expressed as p-values and the overlap is expressed as a probability (this is because the bell curve is used in probability theory as well. Based on the curve of a particular dataset, you can get the likelihood of getting a particular result).
Now remember what I said earlier about the definition of the p-value being probability of getting a result that is equal or higher than the result that you got , assuming the null hypothesis (which is the opposite of your hypothesis) is true. Now, how this is computed is this: in this experiment where you have two conditions (experiment and control), your control is the null hypothesis and the experimental condition is the alternative hypothesis (or just the hypothesis you'd put down in your ERA). The p-value is basically said how likely are you going to get the result that you got, if the control condition is true (that is, there's no effect). A p-value of .05 basically says that there is 5% chance that you got that particular result, given that there's actually no real effect of the treatment. Basically what we are concluding that it is unlikely that this is actually the case (that there's no effect), and therefore the treatment must be doing something (although the degree of what it is doing can't be determined using tests of significance along).
Where t-tests come along on this is that you use a Student's t-test to come up with a p-value. It's not the only way to come up with a p-value and you can only use it when you have two conditions that you are comparing. Nonetheless, it works by the principle I've stated above.
This is probably enough for you to get an handle on what a p-value is (and you're probably skipped to this part anyway). It'd definitely beyond the standard that you'll need for VCE, but understanding some of the background of Student t-tests should help you understand why we use it.