# Help needed on what Statistical Test to conduct for this Study

#### maxwell294

##### New Member
Hello Everybody!

I am an undergraduate, and I need some help on what statistical test to conduct for a research project I am doing. I am fairly new to statistical tests (only took AP Statistics in High school), and the resources my advisor has pointed me towards didn't seem to have anything that worked, so any help pointing me in the right direction would be highly appreciated.

My study attempts to test the comparative effects of the presence of photosynthetic organisms macrophytes and algae on the growth of E. Coli. To do this, we are introducing algae and macrophytes into vials of a growth medium as shown in the picture attached here.

Because we don't know the relative proportions of algae and macrophytes in the environment, we came up with a system to create multiple blocks with different proportions of algae and macrophytes in order to randomize the proportions. Thus, if one photosynthetic organism leads to an increased number of bacterial growth in the majority of the five blocks, we know that that particular photosynthetic organism leads to more bacterial growth, regardless of the proportion of macrophytes to algae in the environment. For example, as you can see in the figure, the first block has two vials with only macrophytes in one and only algae in the other at a proportion of 1 macrophyte to 5 algae. Then in the vial with combined algae and macrophytes, we retain those proportions when we combine (i.e. there is a mixture of 1 part macrophyte to 5 parts algae in the same vial). We then switch the proportions in each subsequent block — for example, the second block has a vial with only 2 parts macrophyte, a vial with only 4 parts algae, and a vial with a mixture of the 2 parts macrophyte and 4 parts algae.

The trouble is, we aren't able to figure out what statistical test to conduct for this study design we came up with. We were thinking a 2 part ANCOVA could work but weren't sure? Any help on what statistical test we could conduct to verify our results, or if you have any idea on how we can modify our study system to control for the unknown proportions of algae and macrophytes in the environment without doing it in this block system would be very helpful!

Thanks a lot!

#### fed2

##### Active Member
first draw a plot, then if you see any interesting trends run a stats test to confirm.

#### Karabiner

##### TS Contributor
first draw a plot, then if you see any interesting trends run a stats test to confirm.
Beg your pardon, but that's exactely the strategy which steadily produces tons of
unreplicable study results. To eyeball data and then tailor tests to "confirm"
interesting-looking patterns makes it impossible to distinguish between chance
results from random noise and real effects. Eyballing means to perfom a huge
number of implicit "tests" - how is this accounted for in the more standardized
statistical analyses?

With kind regards

Karabiner

#### Karabiner

##### TS Contributor
we came up with a system to create multiple blocks with different proportions of algae and macrophytes in order to randomize the proportions. Thus, if one photosynthetic organism leads to an increased number of bacterial growth in the majority of the five blocks, we know that that particular photosynthetic organism leads to more bacterial growth, regardless of the proportion of macrophytes to algae in the environment.
Your picture shows 5*4 vials. Have the 4 vials in 1 block to be considered as dependent observations,
so that wie can say "there are n=5 'subjects', and each of these 'subjects' was measured under 4 different
conditions"? Or are the vials within one block perfectely independent from each other, so essentially
we have a group of n=5 control subjects, n=5 algae-only treated subjects etc.?

With kind regards

Karabiner

#### fed2

##### Active Member
I never promised you a rose garden....

unreplicable study results
repeating is someone else's problem. you need to get this thing published right? You can go back to being a good boy tommorrow..... Anyway, if your this far down the line on running the experiment without already knowing the analysis method, the ship has sort of already sailed you know.

#### Karabiner

##### TS Contributor
you need to get this thing published right?
Is there such a need? It was only mentioned that this was a undergraduate project. And I do not know
why the OP should be instructed how to produce crap, as long as s/he might perhaps be able to make a
proper analysis, with the same or less amount of effort.
..... Anyway, if your this far down the line on running the experiment without already knowing the analysis method, the ship has sort of already sailed you know.
Yes, preferably the statistical analyis should be determined during the design stage. And as far as
I understand, that is what they are currently in?

With kind regards

Karabiner

#### fed2

##### Active Member
make a proper analysis
All of this properness is not without its cost, depending how proper you want to be. If you were to do 'GLP' type things on every study, it would grind things to a halt, just on the paperwork alone. Not to mention the cost. That's why mostly people don't do that. Who's going to pay for a statistician, not to mention an independent QA department? To everything a season ya knw.