How the hell am I going to achieve this?

[theresnowayicandothis]

I've got an essay due in 10 days for uni. If I don't pass it, I can't do my final year at uni. It's a psychology report (2nd year level), I need to do some stats on SPSS and G*power and stuff to do with correlation. Also something about inferential and descriptive statistics, and assumptions.

I've had a shit year, a lot of adverse circumstances, I hardly went to any of the psychology lectures and I'm bewildered by most of what I'm reading.

I have higher maths, but that was in 2009, and I only vaguely remember doing statistics. I've worked out what a standard deviation is, what normal distribution is, what a z score means and what correlation is - I think. Mainly from memory from school!

Uni are aware and extremely supportive and have told me they believe I can do this. I've been given 6 hours tomorrow and the next day in my mentor's office to sit and get on with it, to eliminate chance of me sitting panicking and not getting anything done.

I realistically want it done by Friday, but I've got until the 27th. Uni said if essential I could get an extension for a further week, if really needed, they'd support that.

How the hell do I do this? Are there websites that explain this in a very simple fashion?

Lots of people like the Andy Field statistics stuff. He's got a website with some free resources.

Is it worth revisiting the lectures? Are they recorded?

Youtube videos are your friend

Statistics without maths is a really good book, talks all this through in a very straightforward manner. Statistics by Jim is a good website too.

Stats are basically all about working out whether there are genuine patterns in data or the differences are just random. Descriptive stats are just talking about your means, standard deviations, etc. Inferential stats are the ones that tell you whether your hypothesis has been supported or not.

Suppose you're studying memory. You want to conduct a memory test on two groups - undergraduates and postgraduates. Your hypothesis might be that postgrads would perform better. You would then test, e.g, 100 people from each group. Your descriptive stats would show you very quickly if your hypothesis was broadly right or wrong. If the mean score among undergrads was 20 and postgrads was 80 then you know you're on the right track. But what if the mean score among undergrads was 49 and postgrads was 51? Those differences might have just been random. That's where your inferential statistics come in - you choose the appropriate test (in this example a t-test or Mann Whitney), run the test on SPSS and it will tell you whether your differences are statistically significant or not. Voila! All the different stats tests are for slightly different types of data.

Julie Pallant’s SPSS survival guide is brilliant. Talks you through exactly how to do statistics in SPSS. Really good luck!

There is often study support available to students during the summer holiday. There are usually staff, perhaps based in the library (?) - ask at Student Services/Student Hub - that you can make an appointment to see and they can help with, amongst other things, stats support. They usually support Masters students with their dissertations during the summer but they will also support undergrads. Your university is probably subscribed to LinkedIn Learning and I have found lots of their training to be excellent (I did training in PowerBi and Excel). I often recommend it to my students. SPSS is basically fat Excel, so if you can use that, you know about 70% of SPSS.

Finally, go back to your Blackboard/Moodle/Course website. All the learning materials should be there, including the PowerPoints and reading materials, and hopefully the lecture recordings. You will also have available (heaven forbid!) the module text - in my day it was Coolican which is a good source of guidance. You need to adopt a problem focused coping strategy here (can you tell I’m an academic psychologist yet? 😂). In your shoes, I’d step back from the essay for a day or two and learn how to do the stats first. It will give you a lot more confidence and actually, once you get the hang of them, stats are pretty exciting.

Ok so an update (I have read this thread through several times over, it's been really helpful).

I've spent all day Tues, Wed, Thurs and today at this. I've managed to write most of the background stuff, introduction etc.

I've spent all day trying to do G*power and SPSS.

I'm now beyond confused. When I'm doing G*power, I'm being told the sample size I need is much bigger than the one I've actually got.

More confusingly though, I was told to generate sub scaled scores, look for normality/skewedness. At least one set of data has come back skewed. I have no idea what that means or what might be causing it. When I did it originally I used the initial data, that came back with everything skewed. When I changed it, I'm still getting skewed but only on one set of data.

From my understanding if it's skewed I can't do anything more with it? Because it would mean all my results are unreliable? (Same with the cronbach alpha thing). It seems odd that everything would come back as completely unexpected results that are very hard to interpret. Makes me think I've gone wrong somewhere somehow.

Have emailed the department - on AL until after the resit date and no-one to contact unless urgent, I'm not sure if this is!

The only reason it doesnt add up is because if it's not normally distributed I need to use stats that were taught end of semester 2.... this report was originally done in semester 1... so I'm convinced that can't be right, to do an exam on stuff that wasn't actually taught.

Have you used Andy field?

YouTube him, he also has a webpage statistics hel-p I think it is and his book of course. You can probably follow a YouTube video to run and report a basic correlation.

If your data is skewed you can transform it (see Andy field)

If not then you can use a non parametric correlation like spearman's rho rather than Pearson's r (just check a different box in the correlation procedure in SPSS)

If your sample is too small going by g*power (how small are we talking?) Then at this point it just goes in the limitations with a discussion about type 1 and type 2 error and more research needed for conclusive and generalisable results. Unless you can magic up a larger sample quickly

When I say 'if not' I mean if you don't want to transform (Andy field provides a discussion of reasons and implications of doing so)

It’s ok if data is skewed, just means you need to use non parametric test eg you would use Mann Whitney instead of t test. In reality sometimes data will be skewed!

Im afraid I’ve never used GPower so can’t comment on that. But normally if your sample size is too small so inadequate power you would just write that up as a limitation of the project.

Not sure if anyone’s mentioned Laerd Statistics. I used them during my psych MSc and it was so helpful. There’s a small monthly fee, but definitely worth it: https://statistics.laerd.com/

I had never really studied stats before, but found it really interesting once I got the hang of it.

As above, Andy Field was great too. Good luck, you can do it.

That’s a relief re skewed, I panicked as I figured everything would come back as being normal (ie easy to understand), so I thought I must be going wrong somewhere. I think I’ve actually done more than I thought then. Thanks so so much.

Ask stats lecturer or someone similar to recommend a text book

You need to speak to your tutor.

I owe my MsC to Andy Field!! Thoroughly recommend him.

To reassure you - every single person I knew on my undergrad and postgraduate psychology courses struggled with stats and SPSS. It’s not easy and it sounds like you’re working really well through it

It's your sample size. With a larger sample you would probably find the distribution becomes more normal.

That doesn’t fit with my textbook/lectures notes though, they all say a sample size of 30 is treated as large … I’ve over 3 times that. I’ve done absolutely everything they’ve said to, followed every single instruction - deleted outliers, etc - and all I’m getting is that nothing makes sense - all data is skewed and unreliable and inconsistent. I can’t see how that could be the case for an undergraduate degree.

im also beginning to worry that this is going to be seen by one of my lecturers and treated as collusion, so may ask for this thread to be deleted. I’ve never felt so frustrated and upset over an assignment before - if I fail it I’m out of uni, if I’m out of uni I’m out of a job, and a home, and in a panic.

When you do SPSS I thought, like stata, there was a command line window that shows the commands you're using. If you copy & paste those commands some of us could advise.

Can you generate & copy here a histogram plot of your data points? If it's non-parametric data then Wilcoxon sign rank test (Mann Witney) would be a better comparison for 'average difference' than t-test, Kruskal-Wallis ANOVA not regular ANOVA, does gpower do non-parametric?

I wondered why you couldn't use spearman Rho for correlation if the data aren't sufficiently Gaussian/normal distribution. Spearman is a default option in SPSS menus and interpretted like Pearson, so easy to use.

I love stata & R is alright but SPSS is devil's work. It's big merit is being menu driven for huge datasets.

I've done a spearman's (which broadly matches what I expected) but won't let me do a Mann Whitney due to my data (likert scale). I've managed to get the correlations now and have been able to comment on them so hopefully that's good enough to not fail it completely. Have checked marking guide and lowest passing grade allows for glaring errors.

I just need to pad out the intro and discussion with references, do appendices and then referencing list.

lowest passing grade allows for glaring errors

That's good news, to relieve the pressure on you a bit.

For the record, I got an amazingly good score by mostly discussing what the errors were, while most of my cohort tried to gloss over the issues and form a conclusion one way of the other.

You have got this. And it is totally valid to discuss what you are doing in real life and on a message board. You are using the answers to inform your thinking. It's not as if you are asking a question then using the answer verbatim.

For extra widows, though, you might consider including some of the questions.answers from chatgpt and other sources as an appendix. In the spirit of openness.

All the best.