Statistics How To

T Test in Excel: Easy Steps with Video

Microsoft Excel for Statistics > T Test in Excel

How to do a T Test in Excel — Overview

t test in excelThe t test is a way to tell if the difference between before and after results is significant or if those results could have happened by chance. For example, a drug manufacturer might test a new drug and compare the before and after results to see if the drug was effective. It’s mostly used to test if means are different. The larger the t-value, the larger the difference in the two samples.

You have three options in the Data Analysis Toolpak (How to load the Data Analysis Toolpak) for a t test in Excel. Read below if you aren’t sure which test to choose or skip to the section you need:

  1. Paired Two Sample for means in Excel.
  2. Two Sample T test in Excel assuming Equal Variances.
  3. Two-sample T test in Excel assuming Unequal variances.


T-testing is used in hypothesis testing, when you are deciding if you should support or reject a null hypothesis. Which t test in Excel you use depends mostly on what type of data you have. If your data has two sets of observations from the same group (for example, medical testing before and after a drug is administered to the same group of people), you would use the paired two sample for means. Otherwise, use a two sample test for variances.

Paired Two Sample For Means is used when your sample observations are naturally paired. The usual reason for performing this test is when you are testing the same group twice. For example, if you are testing a new drug, you’ll want to compare the sample before and after they take the drug to see if the results are different. This particular t test in Excel used a paired two-sample test to determine if the before and after observations are likely to have been derived from distributions with equal population means.

The other two tests are used when you have different groups (i.e. you aren’t testing one group twice over time). The Two-Sample assuming Equal Variances test is used when you know (either through the question or you have analyzed the variance in the data) that the variances are the same. The Two-Sample assuming UNequal Variances test is used when either:

  1. You know the variances are not the same.
  2. You do not know if the variances are the same or not.

In most cases, you don’t know if the variances are equal or not, so you would use the Two-Sample assuming UNequal Variances test.

Back To Top

Paired Two Sample for means in Excel 2013.

Two sample t test for means in Excel 2013
Watch the video or read the steps below:

Two sample t test in Excel for means: Overview

A two sample t test for means is normally used when you are testing twice on the same subject. For example, in a medical trial you might want to know if a particular medicine is effective so you test patients before the medication is administered and after. The t-test can tell you if the results from the trial have statistical significance (i.e. it worked) or if the results probably occurred by chance.

Two sample t test for means in Excel: Steps

Step 1: Type your data into Excel. As the two sample t test paired two sample for means is usually used for “before” and “after” data, you’ll probably have three columns: the first column for the subject identifier (i.e. a name or a number), the second column for the Before results and the third column for the After Results.

Step 2: State your null hypothesis (How to state the null hypothesis). For example, your null hypothesis might be that the means are the same.

Step 3: Click the “Data” tab and then click “Data analysis.” If you don’t see the Data Analysis option, load the Data Analysis Toolpak.

Step 4: Click “t test paired two sample for means” from the options window then click “OK.”

Step 5: Click the “Variable 1 Range” box and then select your first variable list (usually the Before list).

Step 6: Click the “Variable 2 Range” box and then select your second variable list (usually the After list).

Step 7: Type a number into the Hypothesized Mean Difference box. For example, if your null hypothesis stated that there was no difference between the means, enter “0.” Otherwise, if you are hypothesizing there is a difference, type that difference into the box.

Step 8: Check the “Labels” box if you have included labels.

Step 9: Type an alpha level into the alpha level box. An alpha level of 0.05, or 5%, is standard in hypothesis testing so if you aren’t sure what alpha level you need, leave this at 0.05.

Step 10: Click the Output Range box and select an area to the right of your data.

Step 11: Click “OK.”

Reading The Results from the two sample t Test for means in Excel 2013

Your results will include a lot of data, some that’s obvious (like the number of data items). But when you run a t-test you’re really only looking for two things: t-scores and alpha levels.

Step 1: Compare the alpha level you chose (i.e. 0.05) to the p-value in the output. If the p-value in the output is smaller than the alpha level you chose, reject the null hypothesis.

Step 2: Compare the t-critical value in the output with the t-value. If the t-value is smaller than the t-critical value, reject the null hypothesis. There are two t-critical values, one-tail and two-tail. If you aren’t sure if you have a one-tailed test or a two-tailed test, always compare the t-value to the two-tail t critical value.

In order to fully reject the null hypothesis, use both values (p and t) in combination. In other words, if you think you might reject the null based on a small t-value, but your p-value is large, then don’t reject the null.

Check out our YouTube channel for more Excel stats help and tips!
Back To Top

Two Sample t test assuming Equal Variances.

Watch the video or read the steps below:

Two sample t test Assuming Equal variances in Excel 2013: Overview

Two sample t test Assuming Equal variances in Excel 2013A two sample t test assuming equal variances is used to test data to see if there is statistical significance or if the results may have occurred randomly. This is one of three t tests available in Excel and of the three, it’s the one least likely to be used. Why? In the vast majority of cases in hypothesis testing, you don’t know the population variances. This test should only be used if you have been explicitly informed that the population variances are equal. If you don’t have this information, you should be running the other t test (Two sample t test Assuming Unequal variances).

How to do a Two sample t test in Excel Assuming Equal variances: Steps

Step 1: Type your data into a worksheet. Generally, you’ll have a list in one column and another list in a second column. The t-test will allow you to compare the means from these two columns.

Step 2: Write the null hypothesis (How to state the null hypothesis). For example, your null hypothesis might be that the means are different by a certain amount.

Step 3: Click the “Data” tab and then click “Data analysis.” If you don’t see the Data Analysis button on the toolbar (to the far right of the Data tab), load the Data Analysis Toolpak.

Step 4: Click “t test two sample Assuming Equal variances ” from the options window then click “OK.”

Step 5: Click the “Variable 1 Range” box and then select your first data list.

Step 6: Click the “Variable 2 Range” box and then select your second data list.

Step 7: Type a number into the Hypothesized Mean Difference box. For example, if your null hypothesis stated that there was no difference between the means, type “0.”

Step 8: Check the “Labels” box (you’ll usually want to include labels so you can easily compare the two sets of data).

Step 9: Type an alpha level into the alpha level box. If you don’t know what alpha level you should be using, leave it at 0.05.

Step 10: Click the Output Range box and select an area for your output.

Step 11: Click the “OK” button.

Reading The Results from two sample t test Assuming Equal variances in Excel 2013

Step 1: Compare the alpha level you types into the two sample t test Assuming Equal variances window (i.e. 0.05) to the alpha level listed in the output on the worksheet. If the alpha level in the output is larger than the alpha level you chose, you will be unable to reject the null hypothesis.

Step 2: Compare the t-critical value in the output on the worksheet with the t-value listed. If the t-value is larger than the t-critical value, you can reject the null hypothesis. There are two t-critical values, one-tail and two-tail. If you aren’t sure if you have a one-tailed test or a two-tailed test, always compare the t-value to the two-tail t critical value.

Check out our YouTube channel for more stats help and tips!

Two-sample t test assuming Unequal variances.

Two Sample T Test Unequal Variance Excel 2013
Watch the video or read the steps below:

Two Sample T Test in Excel Unequal Variances: Overview

A two sample t test assuming unequal variances is the most common type of t test in Excel 2013. You have three options in Excel for t tests: assuming equal variances, assuming unequal variances and a paired two sample. The paired two sample for means in Excel is generally used if you have a sample you’re testing twice (i.e. a “Before” and an “After”) while the two sample test assuming equal variances is only used on the very rare occasion you know the population variance.

Two sample t test in Excel 2013 for unequal variance: Steps

Step 1: Type your data into a worksheet in two columns.

Step 2: State your null hypothesis (How to state the null hypothesis) (i.e. the means for both sets of data are the same).

Step 3: Click “Data” and then click “Data analysis.” If you don’t see Data Analysis, load the Data Analysis Toolpak.

Step 4: Click “t test two sample Assuming unequal variances ” and then click the “OK” button. This will open the t test two sample Assuming unequal variances dialog box.

Step 5: Type the location for your first set of data into the “Variable 1” box.

Step 6: Type the location for your first set of data into the “Variable 2” box.

Step 7: Type a number into the Hypothesized Mean Difference box. The hypothesized mean difference should have been stated when you wrote your null hypothesis. For example, if you think the means are the same then the hypothesized mean difference is 0.

Step 8: Check the “Labels” box (assuming you included labels for your data, which is always a good idea).

Step 9: Click the “Alpha level” box and then type an alpha level. the default is 0.05, which is a standard alpha level for these tests.

Step 10: Click the “Output Range” box and select an area for your output.

Step 11: Click the “OK” button.

Reading The Results from two sample t test unequal variance Excel 2013

  1. Reject the null hypothesis if the alpha level in the output is smaller than your stated alpha level. For example, if the alpha level in the output is 0.03 and your alpha level from Step 9 was 0.05, you can reject the null hypothesis.
  2. Compare the t-value with the t-critical value. If the t-value is smaller than the t-critical value, reject the null hypothesis. There are two t-critical values — one for a one-tailed test and one for a two-tailed test. If you don’t know if you have a one or two tailed test, use the two tailed test figure ( How to tell if you have a one-tailed test or a two-tailed test).

Back To Top

Check out our YouTube channel for more Excel for statistics help and tips!

If you prefer an online interactive environment to learn R and statistics, this free R Tutorial by Datacamp is a great way to get started. If you're are somewhat comfortable with R and are interested in going deeper into Statistics, try this Statistics with R track.

Comments are now closed for this post. Need help or want to post a correction? Please post a comment on our Facebook page and I'll do my best to help!
T Test in Excel: Easy Steps with Video was last modified: October 21st, 2017 by Andale

34 thoughts on “T Test in Excel: Easy Steps with Video

  1. Andale Post author

    Hello, Darin,
    Glad you like the videos. And thanks for catching the typo…yes, it should say UNequal. I’ve changed it.
    Best,
    Stephanie

  2. Bryan

    Hi,

    I’ve got a question regarding one of my statistics, im running a statistics in determining resolution effects on average SST and i want to compare how significant their changes are. Which statistics t-test would I be running? I assume its the t-test 2 sample for means?

  3. Laura

    Regarding the reading results for two-sample assuming unequal variances, the video said if the t-value was greater than the alpha value (0.05) you accept the null hypothesis which is what I got for my results, however further down in the writing it says compare the t-value with t critical value and with my results it shows that the critical value is greater than my t-value so I reject the null hypothesis. Am I missing something or have I done something wrong? Could you try explain?

  4. Elio

    Hello Stephanie.
    thank you for these helpful videos. Can you help me with a simple problem? I have two sets of samples: two different genotypes of tomato, one is the control and the other is the mutant. For the character I’m studying (hypocotyl length), the mutant has a higher mean. I want to say if this difference is statistically significant with a T test. I performed a F test which told me that the variance is equal (F<F critical) but I don't know if it is correct to use one tail or two. Can you help me?
    Thank you in advance.
    Elio

  5. Andale Post author

    You need to specify one tail or two BEFORE you do your hypothesis test. It’s based on your hypothesis. For example, if you’re just looking to see if the mean is “different” then it would be two tail. It looks like you are hypothesizing that your mean is just greater so that would be a one tail test.

  6. Uthra Venkatram

    Hello Stephanie,

    I started loving your website. Simply superb!!

    Few comments on this t-test: (I have extracted results from your notes and mentioned my comments below each results)

    Reading The Results from the two sample t Test for means in Excel 2013:

    Step 2: Compare the t-critical value in the output with the t-value. If the t-value is smaller than the t-critical value,reject the null hypothesis.

    My comment: (Your video also supports the below, however, in the step 2 it has been mentioned the other way. Kindly confirm)

    If Then
    t-Test statistic > critical value (i.e. t> tcrit) Reject the null hypothesis
    t-Test statistic < critical value (i.e. t< tcrit) Cannot Reject the null hypothesis
    t-Test p value Alpha Cannot Reject the null hypothesis

    Reading The Results from two sample t test Assuming Equal variances in Excel 2013

    Step 2: Compare the t-critical value in the output on the worksheet with the t-value listed. If the t-value is larger than the t-critical value, you cannot reject the null hypothesis.

    My comment: (Your video also supports the below, however, in the step 2 it has been mentioned the other way. Kindly confirm)

    If Then
    t-Test statistic > critical value (i.e. t> tcrit) Reject the null hypothesis
    t-Test statistic < critical value (i.e. t< tcrit) Cannot Reject the null hypothesis
    t-Test p value Alpha Cannot Reject the null hypothesis

    Reading The Results from two sample t test unequal variance Excel 2013

    Step2: Compare the t-value with the t-critical value. If the t-value is smaller than the t-critical value, reject the null hypothesis. There are two t-critical values — one for a one-tailed test and one for a two-tailed test. If you don’t know if you have a one or two tailed test, use the two tailed test figure ( How to tell if you have a one-tailed test or a two-tailed test).

    My comment: (Your video also supports the below, however, in the step 2 it has been mentioned the other way. Kindly confirm). Also the df is 22 (that is 24-2) but in the excel sheet it has been mentioned as 21. Please tell if there is any reason for the same.

    If Then
    t-Test statistic > critical value (i.e. t> tcrit) Reject the null hypothesis
    t-Test statistic < critical value (i.e. t< tcrit) Cannot Reject the null hypothesis
    t-Test p value Alpha Cannot Reject the null hypothesis

    Please be so kind to clarify the points above.

    Thanks a lot!

    Best Regards
    Uthra Venkatram

  7. Andale Post author

    Hello, Uthra,
    Thanks for your comments! I went back, watched the videos and read the Step 2s. I had a small typo in there that was confusing the matter. I typed cannot instead of can.
    To confirm, if you want to reject H0, your t-value must fall into the tail area that is cut off by the t-critical value. In other words, your t-value must be larger than the t-critical value if you want to reject H0.
    Regards,
    Stephanie

  8. Andale Post author

    Hi, Laura,
    There was a typo in the written steps that I corrected so that may have made it confusing. If your t-value falls into the rejection region, reject H0. In other words, if your t-value is larger than your t-critical, reject H0.
    The p-value in the video was greater than the alpha level, so we couldn’t reject the null. I couldn’t find a point in the video where I compared the t-value to the alpha level? If you can find the time (mm:ss) that would help and I’ll take another look.

    Stephanie

  9. Mohannad

    hi
    I like to use t test in order to know the best value for the variable. Any test I can apply it ??
    If possible, re-explain to me the comparison process resulting from the t test.

  10. sal

    Dear Ms Stephanie. Thanks for the videos. I have a doubt in the t-test for Paired Two Sample for means . In this video t stat < t critical but p value is greater than 0.5, so it the hypothesis still rejected as null. Kindly advise. Thanks & Regards

  11. Andale Post author

    In order to fully reject the null hypothesis, use both values (p and t) in combination. In other words, if you think you might reject the null based on a small t-value, but your p-value is large, then don’t reject the null.

  12. Sarah

    Money spent AFTER financial advice is lower than BEFORE the advice (330 subjects tested twice, unequal variances).

    Ho (null) is: the money spent after advice is lower than the before advice.
    (lower: unspecified number)
    H1: (alter): there is no significant difference between (before & after) money spent.

    i.e. advice didn’t significantly affect money expenditures.

    How can I write the hypothesized mean difference?

    Excel doesn’t accept 0< or <0.

    PLEASE HELP

  13. Andale Post author

    Hi, Sarah,
    If your null is that there is a difference, then — seeing as the null is the accepted fact — you *must* put something in here. After all, if you’re stating that it’s generally accepted there is a difference, then that null hypothesis must come from somewhere, right? Take a look at prior research (like journal articles) to find out where that fact came from, then use that.
    On the other hand, if you really have no idea, and you’re just guessing that there might be a difference (i.e. you have no facts to back it up), I would just run a test with a hypothesized mean difference of 0. That way, Excel can tell you if there IS a difference or not.

  14. Elliot

    Hi Stephanie,

    Your short YouTube videos are very helpful.

    I would just like to ask you a question. My final year undergraduate thesis is soon to be submitted and I need to identify the correct statistical test to analyse my data.

    I wish to compare the predation rate of a population of cats with bells vs. the predation rate of a population of cats that do not wear bells.

    Would the correct statistical test to use be a t-Test: two-sample assuming equal variances?

    Please let me know,

    Thanks,

    Elliot

  15. Anais

    Hi,
    Thank you for your website and videos !
    I have a question, I did a t-Test: Paired Two Sample for Means on Excel and the results are :
    for the one tail : there is a difference (t-critical t stat)

    What does that mean ?
    Does this give us an indication on the direction of the difference?

    An other question, still on a t-Test: Paired Two Sample for Means on Excel, if the result is a difference (for one-tail and two-tail) : how can we know the direction of this difference ?

    Thank you very much for your help :) !!

  16. Andale Post author

    “An other question, still on a t-Test: Paired Two Sample for Means on Excel, if the result is a difference (for one-tail and two-tail) : how can we know the direction of this difference ?” The two tail does not have a direction (by definition, it could be either way — that’s why it’s two tailed).

    The paired test is testing for pairwise differences, so I don’t know why Excel would give you a 1- tail sig unless you specified a one tailed test. You would have had to specify it in your hypothesis (what did your hypothesis say?). Could you possibly post the exact output? What were your inputs? (e.g. for the hypothesized mean difference and such…).

  17. Ian Clements

    Hi Stephanie,

    Thanks for your videos; helped to bring my rusty stats back to some extent.

    I would appreciate your advice on my usage of determining whether I have meaningful correlations between any one of my many independent variables, with cases >10, and the dependent ones (one at a time).

    At present, I use a combined IF statement to first check if the T-TEST is below 5%; and then if the correlation is more than 0.05, using, for example:
    IF(T.TEST(E$4:E$60,$SO$4:$SO$60,2,1)$B$64,SLOPE($SO$4:$SO$60,E$4:E$60),””),””)),””)
    And, if both those criteria are met, then generate the ‘a’ of ax+b to give me the strength of the relationship.

    Does all that make sense please?

  18. Andale Post author

    Hi, Ian,
    When you say “if the correlation is more than 0.05,” that doesn’t make much sense. Correlation has nothing to do with the T-test. Your data could be highly correlated and not significant at all (that’s called a “spurious correlation”). I *think* what you are actually trying to do is test for significance between your independent variables and dependent ones. But what exactly are you testing for? For example, if you suspect the mean values of the measurement variable are the same in two groups, run a two sample t test. You might want to check out this article to see if you are running the correct test.

  19. Ian Clements

    Hi Stephanie,
    Thanks for getting back to me.

    I had a typo; I meant corr>0.5. I realise this and the t-test are different.

    As I understand things, maybe wrongly, is that both the corr. and the t-test boundaries are judgement calls. For the t-test, the 5% limit is used on the basis of accumulated experience that with 5%, then there’s a 1 in 20 chance of the difference being due to chance; and this has to balanced against that of missing a real difference if this is not considered enough. So the 5% is thought to be a good balance between missing a real difference and thinking one is found when there isn’t one.

    Now, for correlation, there is the judgement call of what strength of correlation is indicative of a real relationship. Here I am more uncertain; but I have a feeling that the greater the n, then the more any particular correlation is indicative of a ‘real’ correlation and not an accidental one.

    So with even a small n, then a correlation of 1 is nearly always pretty good. Less, then a greater n gives a greater confidence that any particular is one to have a higher confidence in.

    In my case, daily lifestyle tracking in relation to several health metrics, I’m looking to see which of many independent variables (exercise, supplements, nutrition, etc) have an influence on each health metric.

    So I reason (and here I’m hoping you will be patient with me and give guidance) that if (a) there is a t-test result of lower than 5% between my independent variable (say exercise) and my health metric (say trunk fat), then it is possible there is some connection between the two – change the former and it is likely there’ll be a change in the latter.

    Now I further think that if, even if there’s a relationship (as indicated by the t-test <5%), is it strong enough to bother with – if, say, the correlation is 0.1? Should I have a criterion of 0.5?

    Assuming I find that said variables do both have a p0.5, then I thought it reasonable to look at the ax+b relationship and say that the ‘a’ gives a measure of how much the independent variable – exercise – may be influencing the health metric – trunk fat.

    I hope that both clarifies what I’m trying to do and gives you enough idea of how I’m attempting to improve my health to let me know if I’m on the right track. Or perhaps tell me I just need the t-test; or just the corr.?

    Thanks, Ian

  20. Andale Post author

    Hi, Ian,

    It sounds like you are confused with the differences in the procedures. That’s not surprising as what you are trying to do (compare multiple independents and dependents) is a very advanced statistical procedure. You cannot use a t test for this unless you break your data down into part. I would recommend starting with a straightforward correlation for one independent variable and one dependent variable. That will give you any idea of which of your variables are correlated. That said, if you do that you’re going to miss all of the effects between variables. What these might be I don’t know as 1/you don’t say what your dependent variables are and 2/I’m not an expert in exercise science :).

    Start with the simple correlations and see what they give you. That might be enough for your purposes — but it wouldn’t hold up in court ;)

  21. Ian Clements

    Hi Stephanie,

    Thanks for that guidance.

    Yes, I’m aware I’m trying to do something quite sophisticated; and, as you imply, that’s what’s needed.

    I’ve done post-grad stats; and used SPSS for my PhD – but all 40+ years ago. So I know what’s required – but am so rusty that I’m no longer sure how to do it.

    I do do simple correlations – which is where I sought your help. I realise that the greater my n, then greater the probability that any particular r will be indicative of a real relationship. So what I am trying to do – I suspect wrongly – is to use the t-test to give a handle on this. I’ve been daily tracking my lifestyle variables for many years, increasing variables significantly when diagnosed with terminal cancer and given weeks to live – nigh ten years ago. I realise there’s many possible reasons for my survival, but suspect lifestyle has some effect.

    I do use multiple linear regression for some of the analyses – mainly body composition and blood pressure; these are measured daily (n>2,000). And in a sense, the program I use (Statistica) takes care of the stats and how variables interrelate (I use foreword stepping ridge), giving me a clear steer as to what probably helps those metrics.

    However, for such things as cancer and kidney metrics, I can only practically do these sporadically – roughly monthly. So I have far fewer dependent outputs (n 15 – 69), so cannot do the multiple linear regression – as you’ll know, it is best to have n more than 10/variable. So I’m back to simple correlations for these. So I’m somewhat cheekily asking you whether there’s a way to automatically assess the worth of the correlation – which is where I came in, is asking whether using a t-test with the r can give me that.

    I hope you don’t mind this lengthy explanation.

    Kind regards, Ian

  22. Andale Post author

    ” Is there a way to automatically assess the worth of the correlation – ”
    If you are asking if a t-test will assess the worth of a correlation before running the actual correlation, the answer is no. I think what you are asking is “how do i avoid spurious correlations?”. There’s no easy answer, other than establishing causation — which you can’t do with something easy like a t-test :).

  23. Ian Clements

    Thanks Andale. Of course, you’re probably right. But I’m still left with the problem of how to assess the worthiness of a correlation. It seem clear to me, and also from what I’ve read about this, that correlations based on larger numbers are more significant than those based on small numbers; and I thought there’d be an automatic test to check this – even it I was wrong to think it was the t-test.

    I have correlations based on cases as low as 12, but also as large as 3,000+. My gut feeling is that, say, a corr of 0.1 based on 3,000+ is probably more worthwhile than one of, say, 0.5 based on 12. So where there may be a disagreement between the two (I’m attempting to check whether, say, number of steps/day is good for blood pressure [n>3,000] comes out with a +corr of 0.1 makes it good to do even where, say, a -corr of 0.5 is found with an n of 12 for a kidney disease marker. Of course, ideally, the correlations would be in the same direction for both (and often are; but not always).

    So I was hoping for an easy way to automatically check, against a criterion responsive to n, if a correlation could be used.

  24. Andale Post author

    And that is the big problem with correlations…there could be no causation (“worthiness”). As far as I know, there is no way to check if a correlation is worthwhile. It’s a great tool to identify possible trends, but beyond that, you’re not going to get any real statistical significance unless you run separate t tests on your individual pieces. But the problem with that is (like I said before) you have multiple independent and dependent variables, all connected. My advice stands as before — for a simple overview, run correlations on all of your individual pieces (i.e. one independent/one dependent). You can run a t-test in those individual correlations.

    I’m curious where you read that larger numbers are more “significant” correlations. Could you link me? Perhaps I can clear up confusion there. Without running an actual significance test on the results (i.e. a t-test), I don’t think significance has anything to do with it.

  25. Ian Clements

    And I too am confused; but thanks for trying to help.

    I am only here wanting help with the individual pair of arrays – hence, what you have hinted at is just what I think I need.

    Causation/correlation is always difficult. Philosophically, at bottom, there is only ever correlation – the concept of causation is still wrangled about. So I’m happy to go forward with the idea that correlation may be causation, and if I act on this (change my lifestyle) and the changes of the dependent metric continue in the same way, then that’ll do me as causation; and so I repeat.

    As I’ve said, I’ll deal with multiple variables at once using multiple linear regression analysis where I have enough cases. So for body composition and blood pressure (about 2,000), I can do this.

    But for those output metrics only done less frequently, generally about monthly, I only have 12 – 40 end events – I average the lifestyle variables over the intervening times, and then do each one against the dependent metric. This is what I’m seeking some guidance on.

    So is it possible to put some figure on the significance of each correlation, based on the number of events? Would it be helpful to do a t-test on the relationship between the two arrays to see whether they are significantly related, say better than, say 95% chance? Or 5% chance they are not? Which would be best, if either?

    I hope I’m not straining your willingness to help; and I am appreciating this dialogue.

  26. Andale Post author

    A t test will not tell you how strongly the two variables are “significantly related ” but rather, whether you can extend your sample results to a population.

    Take a look at this article: PSU. It outlines how you would use a t test to establish a correlation. But it won’t give you the “strength” of the correlation. That’s what r is for :)

  27. Ian Clements

    Thanks Arndale. You’ve guided me to a better understanding; and given me a great link.

    I was trying too hard over egging what I had to do to ensure I was getting ‘real’ relationships (critical for my management of my three chronic illnesses, partly by daily lifestyle tracking and relating those to biometrics of the illnesses). So I’ll ditch the corr. and just do the t-test.

  28. Annette

    Hi

    I tested 2 compounds on their ability to inhibit an enzyme. I want to compare them on terms of the ability to inhibit this enzyme. Could I use a t-test for this and which one is the best? And how can I best interpret the results, if for example the t-value is bigger than the tcritical value?

  29. Shafiqah Zainudin

    Anyone can help me? I’m doing experimental study right now, about Natural Dye Indicators for Titration Acid-Base and have done my data collection. Then, I want to use T-test for my data analysis calculation to compare the effectiveness of the Natural dye as indicator based on experiment data collection.

    My problem is, how I want to start to use the T-test in Excel? And, what type of T-test I have to choose? Paired two sample for means or two-sample assuming equal or unequal variance?

  30. Andale Post author

    Shafiqah,
    I’m assuming you have two sets of data. Are they connected in some way? For example, if you have before and after, you would use the paired t-test. Do you know the pop variances? If not, you’ll want to use the test for unequal variance. You may find this article helpful: Students T-test.