I was going to turn in my science project without a t-test because I couldn't figure out how o do it. Friends were trying to help me and I still couldn't understand. I thought about RU-vid. Dude really saved my life. Explained everything. Thank you so much
Really clear instruction. Thanks for taking the time to do this tutorial for us. There are many, many statisticians who seem to enjoy making the subject inaccessible. Whoever put the help together for MS Excel is a sadist.
Hello Mr. Grange, My name is Celio, I'm from Brazil. And I spent hours searching for an easy and practical solution in Brazilian sites. And I just saw your video and solve my problem with statistical data analysis. Thank you and the video is great!
We have to do several experiments and labs at school that require to find the p value in excel, and I have used this video several times to help me find it each time. It's so clear and it so easy to understand! I searched for hours trying to find an easy way to understand how to find it and couldn't find one until I found this video. Thank you so much for it! It literally has saved me (and my grade)!
This is extremely simple and helpful. Where I trip up is when all the books say "we reject the null hypothesis" or similar instead of saying "This means these are different significantly" or similar. I reach the end and I'm thinking "What was the hypothesis again?" This is what youtube stats is good for. Thanks!
It was such a great pleasure to watch your video. Wonderful clarity. You would not want to give up your day job but you would be a star with the BBC World Service with your superb communication skills. Many many thanks. Best wishes.
Jim, I am currently doing my dissertation on how carbohydrate intake effects blood glucose. Can I first of all compliment you on the efficient and easy to understand video on t-test's. This is often an area of confusion for my peers (me included). I have just shown this to the majority of the class, all of which will be using to do their t testing along side SPSS. Thank you
Okay I know it's been 9 years but thank you so much for this video! I was freeking out couse my teacher couldent explain it in a way that made sense to me but this was so easy! it kinda helped my final exams a lot
The E-106 means that you have to move your decimal place to the right 106 places. So your P value would be VERY small eg 0.0000000000000000000... ect. Your data is significantly different :) Hope this isnt too late!
I hope you all find this video useful! If you have any questions, please do drop me an email. And if the video has REALLY helped and you want to buy me a beer, please see www.paypal.com/donate/?hosted_button_id=L692FVF8PNUKU
@DarthHako Yes, you can. The exact formula to use changes with whether you intend to test a between-subjects design or a within-subjects design. For a whithin-subjects design, your t-value = D_xy / (sdD_xy / (sqrt(n)). This looks complex, but D_xy refers to the difference between condition x and condition y, sdD_xy refers to the standard deviation of these differences, and sqrt(n) is just the square root of the number of observations you This website explains it better than I can: have.www.unt.edu/rss/class/Jon/ISSS_SC/Module008/isss_m8_introttests/node3.html
Thank you for uploading all these videos. Great work and very helpful. I have watched your video on t-test. At some point you are saying that we need to decide whether we have the same variances between the 2 groups and I believe that you are saying that this is explained in another video. Could you please refer me to that video? Can we find out by using excel? Would this have anything to do with Levene's test? Thanks again, Athina.
You probably know this by now but for others who may not: when you have '#####' in a cell in excel it is often because the cell is not big enough to show all the digits. A solution would either be to expand the cell or to make it stop at fewer decimals
I so appreciated this video. I have no stats background and do not need the "full math" explanation for what I need to do, and this was perfectly clear and concise!
I am doing my bachelor thesis and I am testing 2 trading strategies. One is an active and the other a buy and hold. The active strategy by and sell the stock when you have trading signals from technical analysis and the other just holds the stock. We have the returns and we think we should calculate the monthly returns and then compare the two strategies in a t-test. Is that right? You have both a sample t-test with known or unknown variances but then you also have a paired t-test.
+Per Forsberg It depends what unit of measurement your returns are in, but it sounds OK. Yes, a t-test sounds like it will be appropriate. Your values are coming from different groups, so it would be an independent t-test, not a paired t-test.
Hello Jim! An urgent question: is the result of this test via excel is the t-value or the p-value? And how do I see these two separately, if possible? Many thanks!!! Sharon
Great video, one question: what if my dataset is imbalanced and I have more than 5 times the amount in one vs. The other? I guess since it uses averaged valued it wouldn't matter but, can you do it?
Thank you so much! This was super easy to follow and made it easy to understand even though it's been a while since I took stats. Thank you again! Life saver. :)
Hi, great video! I’m wondering if you could help me. I’m trying to conduct a t-test to see if a lay betting strategy is worthy or not, I have the 1. Prices of each individual bet 2. Prices of winners. 3. Prices of losers 4. Average price won. 5.average price lost, however, I am struggling to know where to put each piece of data and if it’s the correct data to use. Thanks in advance 👍
This has been a great help to me. It is now very easy to get a "p" value from the "t" test. I have developed a "one-tail" test proposal for myopia prevention from your analysis. Thanks!
Thank you very, very much. It's probably a very easy thing for you, but it was actually very confusing for me until I watched your video. I have a question, though, if you'll excuse me: If I already have the means, SD-Variance and such descriptive results, can I still compare the means just by using them without the actual measurements for each, say, individual?
I'm not sure you'll answer my question in time (or at all) but I'd like to do a t-test to compare the mean of a group to the value that correspond to what they would have obtain if they've answered randomly so it would be a t-test comparing a mean based on real data and a pre determined mean, can you do that with excel? (i'm not sure if this was clear enough to understand, I hope so...)
If the variances of each condition are too far apart, does that explain why my P value is like 4.2, even though the one mean is double the other....????
With regards to the final values....does excel provide the value based on the critical value table for a t-test or does the 0.1672 result then need to be referred to the critical value table at 0.05 (df=7) in order to reject the null hypothesis??? Thanks
+Grant Ormerod gspo13@hotmail.com Hi Grant - the result returned by Excel is *not* the t-value, it is the exact p-value. Normally you would check the t-value you obtain with the critical values in a textbook to see whether your p-value is < 0.05. However, as Excel is providing the exact value you need not do this. In this case, your value is p = 0.1672, so your p is not less than 0.05.
Hi, let's say I have data for pre and post tests.... all i need to do now is to just analyse the two groups' post tests right? As in the data that i need to insert into that two columns are the post tests for both groups right?
Sunitha Rachel Hi - no, you would likely need to take into account the pre-test scores, too. This can be done using a factorial ANOVA, with one factor being "time" (i.e., pre & post test), and another factor for your groups (group 1 & group 2). More correctly, you would likely need to conduct an Analysis of Covariance (ANCOVA), which allows you to control for the difference pre-test.
Excellent! My question, once I have a p-value (result of the t-test) what is the next calculation to determine whether improvement has occurred between two groups of data?
Your p-value is essentially a percentage value. (1 - p-value)*100 = percent chance the difference between groups is statistically significant. A p-value ≤ 0.05 is the gold standard in industry as it indicates your group is 6-sigma from the mean (3-sigma to the left & 3-sigma to the right), which means you can be 95+% confident that your groups are statistically different. Your target p-value can be adjusted depending on how confident you want to be in reporting your groups to be different. Check out Khan Academy for more in depth videos on this topic.
Thank you so much for this brilliant video! I have a question, though.. What if you have more than 2 conditions? Would you have to compare all the conditions individually? (I have different subjects, and 6 conditions in my experiment).
Hi - this would require an Analysis of Variance. You can't do multiple t-tests because you inflate the chance of a type-1 error; that is, finding a statistical difference when one doesn't exist in reality.
I tried your method with different tails/types as well as calculated mine in paper, the excel methods ended up with a different value each time T^T the video is good thanks
Remember Excel is just giving you back the exact p-value. When you do the calculations on paper you are likely calculating the exact t-value (which is quite different). That might explain the differences you are getting.
Thanks so much! A question, what if I want to test whether values in an array (they are in fact simulated values) are statistically significant from a particular actual value?
Hi Jim! I have a group of 58 students 22 of which are female and 33 male. I want to test if there is a statistically significant difference between the the internet use perceptions, practices of male and female students. by doing so, i want to check whether or not gender is a factor affecting students internet use. Of the three types of t-test, which type should i use and why. some points about the sample: they are 2nd year students at a university,all from the same faculty of humanities, it is null hypothesis.
Hi. I want to perform a T test. One of the values in the sample is very large. Should I omit it when I calculate mean, standard deviation and variance?
Working on my data analysis now and I already calculated the mean for 2 groups I want to compare but excel tells me I can't just select the calculated mean. So we can only select a list of values from which it will calculate the mean?
What if I had two columns of consumer age and another row of the purchase amount, and I wanted to see if there was a correlation between the two, which t-test would I use
You would use the =CORREL(range1, range2) function, where range1 is your ages and range2 is your purchase amount. This returns a pearson's r value (between -1 and 1)
@JimGrange. Thanks for that Jim, also how would I go about getting a sample of age ranges from my mass of data so I can do another T-test ? I can't seem to find the answer online
I have a question. You mentioned that the resulting value is the probability/p-value. is it also correct to call that the t-value? My worksheet wants me to list the "t-values and probabilities". I thought t-values and probabilities are the same thing.
Hi Jim, Great Video. I have a doubt in conducting a T-test. So I am doing a thesis where I would check for significant differences in pre- and post M&As. So I have the median of 2 year pre- and post- M&A. Would the median difference be the statistical significance? If so, how would I go about with this computation? Look forward to your response. Thanks Marushka
Marushka Nair Hi - yes, it seems you need to do a t-test here. From what you've said, it seems you have two sets of means: one for pre-, and one for post-. So, you'd do a t-test. (I don't know what an M&A is, so if this is another factor, then a t-test might not be appropriate.)
Great video, thank you. 0.00468 does not mean that it is highly significant though. It just means that it is significant. The degree of significance is not possible to imply with the number.
Correct, I mis-spoke. But, a p-value of 0.004 suggests that data are LESS likely under the null hypothesis than with a p-value of 0.04. But broadly I agree with your statement.
Hi Jim, Great video - thanks for posting. At 0:55 you mention that the related / paired t-test is the test to use if the variables come from the same participant. What method should be used if the data is from separate participants? Say each variable is from a different person and nobody contribute to both Condition A and B. Many Thanks.
Jim. Until seeing your video I had assumed that using the =ttest command gave you a t-stat which you then had to compare to a p value table. Can you explain why this is not the case and what the value you refer to as the p value actually is please?
***** Hi. I do not know why that is not the case. You would have to ask the designers of Excel why :) The p-value is the probability of observing a test statistic (not shown by Excel) as extreme (or more so) as the one you have observed, if the null hypothesis is true. So, a low p-value means that there is a low probability of observing a test statistic as extreme as you have if there is no difference in reality. People believe by reverse inference that a low p-value implies there is a greater likelihood the difference in your data is a real, "significant" difference.
Great video! I have a question for you. I'm trying to do a t test for changes in algae cell concentration. I'm comparing the "before" concentration (3 values) and the "after" concentration (3 values). Would that be a repeated measures experiment? I'm collecting from the same culture to count each time. Also, would it be one tailed since I'm assuming the concentration will increase slightly? Thanks!
+rby82589 Hi - yes, that would be repeated measures. 3 values might not be enough, but you could still do the analysis. If you have a directional hypothesis then 1-tailed is often used, although there is a compelling argument to never use them (it all depends on your field, I suppose)!
video really helped me , but i got an answer of 3.856E-91 ive done it like 10 times and ive done it for other parts and it worked out successfully bar this, would this answer be right or is it an error?
+LauraK2k8 It's worked correctly. That is scientific notation. The "E-91" means move your decimal places 91 places to the left. In other words, your p-value is EXTREMELY small :)
@@sam0812 You only have one mean per 2 countries? T-test is not appropriate here. However, in the general case comparing countries is obviously "between-subjects" (or "unrelated") because the means come from different groups. Therefore, type should equal 2.