Hi everyone
Sorry for yet another boring stats question, but I'm trying to help someone in my office and I'm stumped. I've done this before but can't remember how. We are trying to see if there is a sig difference between a certain value over 12 months. I've tried ANOVAs and t-tests but nothing is coming out right, I can't figure out if it's just how I've set the data up or what. There is no grouping variable or anything, just a single value for each month. I want to avoid doing loads of individual t-tests though...
Thanks in advance!

it's difficult to say from the info you're giving. It all depends on the the question you want to answer.
are you looking are somthing increasing or decreasing then you just plot the value against time, do a linear regression get the r-square value. the closer it is to 1, the better it fits a trend curve and the easier it is to say "parameter x increase with time" for example.
you cannot do T-test or anova by comparing a single point to another and so on!

Hi thanks for that Ariba, I knew that myself, but a friend had his viva and was told by his external to get a P value for the data. I though I must have had it wrong so I posted on here but turns out the examiner didn't have a notion of stats! :)

is it not just a correlation?

I tried that sneaks, but it's not working properly, because I only have one value for each month, it's not working. But it was ages ago anyway now, the thesis is submitted and sorted! But thanks for the help everyone!

well done algae! I've been conferencing so missed out on all the forum gossip! does it feel good to submit?

lol, no no it's not me!! I've still got 2 months!! It was a guy in my office, his external told him to get a P value at his viva, but whatever he ended up doing it's done now!! :) But I look forward to seeing that message again in a few months for me!! :D

ah, sorry I haven't had my morning sugar rush!

I have a friend who's recently passed with no corrections - I looked at the stats and they were COMPLETELY wrong! I guess the examiners didn't even look.

Hehe Algaequeen - not long now til we can all congratulate you!

No correction - wow. When you said they were completely wrong Sneaks do you mean this person used the wrong analysis? Or reported the wrong bits/values?

I know somebody who passed with no correction, he'd already published papers from almost his entire thesis.

It's rare to get no correction right? I just want to pass. Can't wait until the PhD is over...

it seems everyone in my department has got no corrections - the pressure!

It was odd, they'd used analysis that appears correct i.e. comparing groups = anova, but the nature of the data meant that it completely violated one of the assumptions. I'm doing the same analysis, so know a lot about it, I've just spent 2 months sorting it. One of the examiners was a stats expert but obviously didn't pick it up.

Sorry to be hijacking the thread, but Sneaks: any chance the violated assumption you referred to was to do with variance? I've actually had instances where my Levene's test yielded significant p values, but I err ignored it as didn't know what to do, and the stats book says in that case it'd only "compromise" the F value, and my supervisor has read almost all of my analysis results and didn't say anything (though he's not a stats expert at all, to the extent that I suspect I may be more familiar with multivariate stats than he is...). Getting bit worried now over the prospect of having to do all the analyses all over again and rewrite everything...

no, it wasn't levene's. I think you can violate levenes, as long as sample sizes are equal, and if they;re not there are ways around it.

It was the nature of how the data was inputted. It violated the assumption of independence of scores i.e each person had more than 1 score in each category.

Thanks Sneaks. Phew!

Have you told your friend what you thought was wrong with his/her stats? It's lucky that the examiners didn't pick it up, but if they submit papers to journals the mistakes would probably be spotted eventually?

No, I won't, although I might forward my study, which uses the same technique on to her andhopefully it will flag it up without me looking like a complete biatch!