Hi all,

I have some data that I want to do statistical analysis on, but I'm not sure which test to do. I'll try to explain in lay terms what I did...

I had some cells, grown on a membrane - 3 sets of cells in total. I gave each of the sets of cells the same treatment, and then determined the level of a chemical called IL-8 both above the membrane, and below the membrane. So I have three results for level of the chemical above the membrane, and three for level of the chemical below the membrane, on set of results per membrane. I want to analyse whether the amount of the chemical above the membrane is different to that below the membrane.

I think I need to use a t-test, but I don't know which sort, so could someone please help? I am pretty crap when it comes to stats, so any advice would be much appreciated :-)

Many, many thanks...

Matt

Update - I'm pretty sure I need a two-sample/independent-sample t-test. But can someone confirm this?

And if so, what about assuming or not assuming equal variances?

a t-test compares the means of 2 groups.

If you want to compare more than 2 groups, then its ANOVA.

Hang on, will read again - I'm used to people, not cells!

ok, so if it were people..

You have 3 people and you tested their IL-8 once above and once below. and you want to compare above/below.

Right - I think its a repeated measures t-test, because the same participants (cells) are in both the above and below conditions. I.e. you are not comparing cells 1,2 and 3 against 4,5 and 6, but you ARE comparing 1,2 and 3 against 1,2 and 3 in a different condition = repeated measures!

What Sneaks said :)

Pretty sure it's a repeated measures t-test. Again, more used to people though. You're testing the difference between the above and below across all the cells. It's when you start looking at the difference between the above and below within and between individual cells that it becomes ANOVA territory.

Can I add more to the question?

In total, I have two independent variables...

1. Concentration of treatment (three groups - 0, 50, and 100).

2. Inhibitor used (none, inhibitor A, inhibitor B)

Each membrane was treated with one of the 9 combinations of concentration/inhibitor, and IL-8 concentration measured above and below the membrane.  Three membranes were performed for each combination of concentration/inhibitor.

All I want to know is whether, for a set combination of treatment/inhibitor, there is a difference between IL-8 above and below the membrane. I'm not interested in looking at IL-8 variation between inhibitors.  I might also like to know whether, for a inhibitor, the IL-8 either above OR below varies significantly with treatment concentration.

From looking online, it seems like I need to do a repeated measures GLM/ANOVA, but can I just do separate paired t tests? And are there any good sites for people with no stats knowledge about how to enter data for these tests and to run them?

Once again, many thanks...

I hate stats :-(

Matt

your data confuses, me I can't get my head around it not being people!

You shouldn't do multiple paired t-tests though - its bad practice, and causes an inflated error rate. Its better to do ANOVA. I think there are a few stats forums out there. To be honest, if its for your PhD then your better off learning exactly what you are doing so you can defend it in your viva. Andy field is the god of statistics and I recommend his book. He also has a website with loads of materials www.statisticshell.com

======= Date Modified 28 Jul 2010 16:01:33 =======
just to add, it sounds like you might need a MANOVA - your two dependent variables being above and below?? and your IVs being the different treatments you say and inhibitors.

Failing that you could do 2 lots of mixed ANOVAs, one for above and one for below. BUT if you have taken more than one sample per cell in each condition e.g. cell 1, at 50 and inhibitor A - measure 10 times, then you have violated the assumption of independece of scores and can't use ANOVA. If you've only made one measurement per cell then its fine.

Is it not just one big mixed ANOVA? I'm getting confused too. The scores above and below the measure being the within and the groups (inhibitor/concentration) being the between?

So a 3 X 3 X 2?

I could be way off though.

but I *think* there are two dependent variables one being the score for the above and one being the score for the below???

Therefore its either one MANOVA or two mixed ANOVAs (i'd prefer the mixed ANOVAs with post hoc tests if it were me!)

Do they count as two seperate DV's though or two measures of the same DV? If that makes any sense. They're both measures of the level of the same chemical, just from two different locations.

good point - I guess that's up to 4matt to decide!