New here and completely lost :/

Turnip

New Member
Hello everyone! I’m a newbie here and to be honest in desperation googled “psychology statistics forum” thinking no way I’d find one and low and behold there this site was!

Bit of background:

I’m finishing the last module of my second degree in psychology. I’m hoping to do my doctorate next in clinical psychology. I did both my degrees by correspondence otherwise you could bet I would have been staying after class daily to pester the prof for extra help. I have never been particularly comfortable with statistics, although I try my best and so far have managed to keep an average of around 68-72% for stats. Not brilliant but not failing. However, my uni has never really gone into proper detail on how to run the stats. Sure, we get taught the glossary of all the terms, and how to write a research report, but as far as saying “if you’re doing this type of research then this is the type of stats test you use, and if you do that type of research then you use that type…” we just haven’t been shown.

So fast forward to now, and I have to do a pretty basic (I think) research report where I have to do a short verbal and numerical test that looks at short-term memory on 15 people of two different age groups (i.e. 20-40 and 50-70 year olds).

I then have to compare the average score of both groups to each other to see if age has any impact on short term memory. When I do the tests on people I have to:

--> Note how many words they remembered correctly.
--> Note any false recollections they may have (so any words put down that were never read aloud to them).

For the numerical test I have to:

--> Record the highest digit string each person got. (And then again compare the average from the older group to the younger group.

--> Then I have to compare the averages of both groups to each other to see if age has an impact on short term memory.

Now, here’s where I’m starting to eat panic. I don’t have a control group since both groups are being experimented on. (I think I got that right?) So would that mean I do a Chi square test of independence? Or a Mann-Whitney U? Or something else entirely like a paired t-Test? I had originally thought a paired t-test but I wasn’t so sure my two age groups being compared against each other were mirror image style.

So sorry for being such a dunce and hope someone clever can point me in the right direction. I have poured over all my old textbooks from the years and spent countless hours on Google but I’m lost and feeling so worried I am going to mess up my whole report by doing this part wrong.

Thanks so much if anyone can help! I'm sorry if I am being really slow on the uptake here! :shakehead

GretaGarbo

Human
if it in both cases are about the number of words (say like 22 out of 30) and the length of the digit string (like 12 out of 20)

Do a z-test and (for both variables) compare the two groups.

Turnip

New Member
Thanks so much Greta for the reply! It seems the more I read up the more lost I become.

For my word test there are 20 words I read out to the participants. They then do free recall and I record how many they get right and how many incorrect ones they throw into the mix.
Then for the numbers the digits keep going up and up til they make an error. Most people recall around 5-9 numbers in a sequence. I then record how many in the sequence they got right. After doing all this I compare the two age groups. The first age group is age 20-40 and the second age 50-70. I am supposed to be able to tell if age has an impact (positively or negatively) on short term memory. So far just by a quick calculator equation the averages of each group seem about the same. So it looks like age has no factor, and if anything it's very marginal.

Given that I compare two groups side by side do I not do the paired t-test? What makes it unsuitable (just want to check my understanding so I grasp why it's not a good idea)?

Turnip

New Member
Oh, I'm not sure if it makes a difference but my sample is only 15 per group. (We were instructed to stick to 15.) I think I remember that if the sample is less than 30 one should do a t-test? I may be way off there though...

GretaGarbo

Human
If there had been a pairing between the person in the older group and the person in the younger group, like if they had been relatives, the the paired t-test would have been relevant. Now, do a independent group t-test.

I should not have said z-test above. I was thinking about the binomial distribution, proportions test, and and the logit model. But this seem to make it unnecessarily difficult. An other possibility is a permutations test. I just said this as info. Do the independent t-test

Turnip

New Member
Thank you so much Greta! That's really helpful. I have been reading up on the independent t-test and it does look like it is the right one! Everything started sinking into place when I was researching that up. I found a great link on youtube by a guy called how2stats (maybe he's even on here?) that really explained it in such a nice easy to understand way that I think I should be able to follow it.

Can I just check a couple points first?

1) Given that I am comparing both verbal and numerical scores between two groups, how do I calculate that? What I mean is, with what I know about t-tests you compare two independent variable means together and get a reading. But for me I have the required two independent variables (i.e. age group 1 and age group 2) but then there are two sub factors underneath each age group variable. I.e a verbal word score mean and a numerical score mean. So I'm a bit lost there. And I'm not sure whether one can actually include the part where I must record any errors that my participants made in the word recall test, or just include it in the write up part, but if I have to include that in the analysis doesn't that complicate things further?

2) am I right in saying the following? I think I am but wanted to double check:

My Ho is: An increase in age has no impact on short term memory.
My H1 is: An increase in age results in a decrease in short term memory.

Thank you again so much! You have helped me so much!!!

GretaGarbo

Human
You investigates verbal and numerical scores separately.
- check the verbal score difference between age groups
- check the numerical score difference between age groups

You could plot verbal score versus numerical score in a scatter plot and mark the groups with different symbols (and compute a correlation coefficient). That could be interesting. But that has nothing to do with the t-test.

My Ho is: An increase in age has no impact on short term memory.
My H1 is: An increase in age results in a decrease in short term memory.
Yes that is correct and that is a one-sided test.

But you can also do a two sided test:

My Ho is: An increase in age has no impact on short term memory.
My H1 is: An increase in age results in a decrease or increase in short term memory.

That would be a safer set up. Imagine for example that if higher age gives more experience and more associations, thus making it more easy to remember. I am not saying this is plausible, just that it is a possibility, and that you can not rule out that possibility. Almost all tests like this one is two sided. (The usual significance level is 5% = 0.05. This corresponds to 95% confidence level.)

Turnip

New Member
Yes that's actually something I subsequently (to writing my last post) decided I would do (two tailed). Purely because I seem to be seeing that older people seem to be performing better at the verbal part of the short term memory test (but not much more, still need to calculate whether its statistically significant or not) compared to the younger group. And vice versa with the numerical. So I thought hang on an second... I should do a two tailed! I am relieved to see I was on the right track!

Out of interest; once I have calculated the two variables (numerical and verbal) can I not do a further independent t-test to compare he two together? If possible I would love to throw in something extra, such as the scatter plot you mentioned, to make it as detailed as possible. Desperate to get an A grade on this module! It would mean me finishing my degree with a First (a distinction).

Well, it's 2am here in London so I think I'm headed to bed. Thanks again for your really helpful advice. I will check your reply in the morning. have a great evening/day wherever you are!

Turnip

New Member
I have run the independent t-Test for the 3 experiments (namely how many words are correctly recalled, how many incorrect words are recalled and the longest number sequence recalled) and I think I have done it correctly. But when I did a t-Test for the last part, i.e. the number recall, I found I got a really dodgy figure of 0.0003. Which to me looks totally wrong as it's below the usual alpha value of 0.05.

I tried to put the tables in here but the formating kept getting messed up so I don't know how to share it with you guys.

Basically I have two age groups. Each age group I tested:
-Correct words recalled (from a list of 20)
-Incorrect words recalled (sometimes we make up words in a recall test)
-Longest digit sequence recalled.

So I first wanted to check for variance equality so ran a "F-Test two sample for variances" for the three experiments:

- Words correctly recalled: P-Value of 0.077217333.
- Words incorrectly recalled: P-Value of 0.19765994
- Longest number sequence recalled: P-Value of 0.045664149 (huh??)

Then I ran an independent t-Test and got:

Words correctly recalled: 0.055236223
Words incorrectly recalled: 0.146249158
Longest number sequence: 0.000312392 (quadruple HUH?????!)

I am completely unsure what I did wrong (or did I?) that would fail my H1 when it is so obvious that the older you get the worse your memory performs, which can be seen in just looking alone at the overall average performance of the two groups side by side where the younger group seemed got about 7-9 digits in their sequence recalled compared to the older group who averaged around 6 digits long. Clearly the older group didn't do as well, so why is my figure 0.0003?

I hope someone can help. I'm at a loss.