How to Perform a T Test in Prism

Essential Statistics

This video walks you through the steps required to undertake a t test in Prism. The video also features tips to help you more efficiently navigate and use Prism.

You will learn how to:

  • Correctly enter your data for a t test
  • Analyze the results of your experiment
  • Format and annotate a graph of your ...

This video walks you through the steps required to undertake a t test in Prism. The video also features tips to help you more efficiently navigate and use Prism.

You will learn how to:

  • Correctly enter your data for a t test
  • Analyze the results of your experiment
  • Format and annotate a graph of your results

This video is part of the Essential Analysis series, presented by Dr James Clark, from the School of Cardiovascular Medicine and Sciences at King’s College London.

Transcript:

Hello, my name is James Clark from King's College London, and I'm going to show you in this short video how to carry out a t-test using GraphPad Prism. Prism allows you to plot many different types of graphs and produce many different statistical analyses, but for this example we're going to choose probably what is the most simple statistical test you would ever need to carry out in scientific research, and that is a student's t-test.

In order to continue and complete a student t-test you'll first need to enter your data into Prism. The key to using Prism is understanding how the data tables work, and that'll be covered in another tutorial, but for the purposes of this walkthrough we need to select a column table and graph from the menu on the left. Once we've selected column we can click create. It is worth remembering, whenever you create a new file in Prism you will create a data table and a connected graph. These will, by default, be called data 1, and the graph will be called data 1.

For this demonstration we're going to relabel the data table t-test, you can do this by double clicking on the word data 1, and you'll notice that the graph also changes its name to t-test. The Prism table, in order to carry out a t-test, is laid out as shown on the screen, with group A, group B, group C, and group D in columns, and the individual replicates of each dataset in the rows. At the moment we have no data so I will enter two sets of data in order to carry out a student t-test.

On the screen we have dataset from a fictitious experiment, where have control values in column A and treated values in column B. These two values represent an n-of-8 study where we have produced eight replicates of the experimental outcome, and to undertake a student's t-test we want to compare the mean and standard error of group A with the mean and standard error of group B in order to determine whether there are any differences between these two experimental groups.

The data shown here are unpaired, in other words, the values in group A are different experimental outcomes to the values in group B. In order to carry out a paired t-test each sample will need the have been measured twice, once at control, or maybe baseline, and once after treatment. If your data are not paired you need to carry out a standards t-test, but if your data are paired it will increase the power of your study to undertake a paired t-test.

It is worth remembering that every time you enter data into a table in Prism it will also create a graph for you, so let's look at that graph now. Graphs can be found under the graphs tab on the left side of the screen. When clicking on the graph tab you will get the change graph type option box, and we can switch through the various types of graphs available to us in Prism. Whilst formatting and detailing annotation of graphs will be the subject of another walkthrough, we can very quickly see the kind of graphs that are available to us in Prism.

Looking at the mean, median and error graphs, we can choose between a column, a dot, a line, a horizontal column, horizontal dot, and a horizontal line. We could, of course, show the individual values in the form of a scatter plot, scatter and bar, before and after plot, another scatter plot, this time horizontal, scatter and bar horizontal, and before and after in a horizontal format.

Version 8 of Prism introduced violin plots, in addition to the box and whisker plots available in previous versions of Prism, and they can be found in the middle option box. You've got the option of floating bars, a box and whiskers, the violin plot, and then the horizontal versions of the same. For the purposes of this demonstration we're just going to choose a standard bar graph.

Looking at the graph, we can already see that our treated group appear to have a higher mean than our control group, but in order to test this statistically we will undertake our t-test. Carrying out statistical tests in Prism is very easy, you've got two options. The first option is, either on the graph or the table you can use the analyze button on the menu. In addition, you can click on the new analysis button in the results section. For the purposes of this I'm going to go to the raw data table and click on the analyze button.

When you click on the analyze button an analyze data option window appears, and you can choose from a selection of statistical tests. For the purposes of this we want to do a t-test, so we look down to the column analyses and click on the first option, which is t-test. On the right side of the window we are asked to choose which datasets we wish to analyze, and you can see here, we've already selected by default the control and treated groups. It's worth remembering that a t-test only allows the testing between two groups. Once we've selected our two groups we can click on the OK button.

Now we have a series of options to choose from in order to undertake the correct test. As discussed previously, these data are unpaired, and you can see from the figure that we are comparing the means of group A with the means of group B, the control and treated groups. If we were to select paired from the option, you can see the figure now shows us that we are directly comparing the data values from group A and group B using a paired t-test. Of course, these data, as I said, are unpaired, so we'll return to the unpaired t-test.

We will assume our data has Gaussian distribution and choose yes from the next choice. We are also going to be undertaking an unpaired t-test and assume that both populations have a similar standard deviation. There are other options we have, including the ability to plot a residuals plot, but in the third options box we can choose whether we have a one-tailed or two-tailed test. For most t-tests we will do a two-tailed t-test, however, if you hypothesize that your data will only move in one direction, in other words, your treated group can only and will only be higher than our control group, we could, in this instance, choose a one-tailed t-test.

We're going to report our differences as treated versus control, and we're going to set our confidence at 95%. This means that if P, as a result of the t-test, is less than 0.05, that's a 95% confidence, we will agree that group A, or our control group, are different from group B, or our treated group. Having left the other choices in their default state we can click OK.

As soon as you have carried out your student t-test your data will appear in the results section on the left-hand side, and you can see here, if I just move the window across, it shows unpaired t-test of t-test, t-test being the name of the table. You can see the table has analyzed our data, and it's compared our treated group with our control group. Our P value is 0.0007, and Prism has told us that our P value summary is three stars.

In this instance, of course, we set our tolerance to 95% confidence, so anything below 0.05 P value is assumed statistically significant. Of course, Prism tells us that, and we can see here that on line 10, significantly different? yes, this is a two-tailed t-test. Below, you can see the outcome of the other parts of the t-test, including confirmation that our N was eight in each of our groups.

So, from the output shown here, we can conclude that group B, or our treated group, is significantly different from group A, or our control group. We can go to our graph now, and we can annotate our graph using the text annotation tool and enter a little star above our treated group, indicating that it is statistically different than the first group.

A nice little feature in Prism is when you have entered a star and you move it, it will snap automatically to the center line of your graph, allowing a nice clean graphical presentation. In recent versions of Prism you can also add a line indicating difference between two groups, you do this using the draw tool, and you select from the menu, lines with text. I can click on line with text, draw a line over the top of my group, and put a star on, indicating significant difference between these two groups.

So, that is a simple t-test, unpaired, between two groups of data in GraphPad Prism.

Show more