Yet, the concept of the Null hypothesis is basically associated with
No, actually it isn't. What gives you that impressions is null hypothesis and student's t-distribution being taught together because a simple 1 sample t test is simple and easy. But the truth is that there are many other sampling distribution besides t-...
Standard paired t-test for two samples y1 and y2 is just the one sample t-test applied on the difference d = y1 - y2.
The same applies when using a normal distribution in a z-test.
(That and the availability of paired t-test in scipy.stats are the reasons why statsmodels currently does not implement a separate paired t-test.)
The first question is what is the appropriate analysis. This is an analysis of change from baseline, specifically whether a proportion change under one experimental condition is greater than the same change under a different experimental condition, presumably performed in separate units.
The most valid analysis is an ANCOVA where the response is log ...
Why not just look it up a T Table or just punch the numbers in excel? I get the elaborate explanation above and good job at it but feels a bit overkill. In excel you can use T.INV. Just that you need degrees of freedom (which for the t distribution is n-1) for the second argument in T.INV(). And then just adjust based on what test it was upper, lower, two ...
You can use the ttest_rel function in scipy.stats:
Calculate the t-test on TWO RELATED samples of scores, a and b. This
is a two-sided test for the null hypothesis that 2 related or repeated
samples have identical average (expected) values.
You test agaist specific alternatives, e.g. changing mean, changing variance, changing higher moments. For each one, one or more tests exist, and they are not necessary based on the $t$-test. The closest one could be the augmented Dickey-Fuller test with a $t$-statistic following a nonstandard distribution under the null hypothesis of presence of a unit root ...
The way to reconcile the opposing views is the following:
1) The t-test does assume normality.
2) The t-test is robust to deviations from said normality.
For #1, we do need normality in order for the test statistic to have the t distribution that we assume. However, #2 kicks in by slowing the data themselves to be non-normal but allowing for the test ...
P-val (raw) is calculated using a df of 11. That's right, the function rounds DOWN. P-val (round) is the same value as the data analysis toolpak, it also rounds UP to 12 as you did. The function T.TEST uses the decimal value of 11.7, so it gives the most correct value.
I used this website to calculate the p values given any given df value and t stat.
The independence requirement for the t-test is not really relevant here, as you have only one students's data (If you had data for more students, that would be more of an issue) ... but there is also an assumption of normal distribution and that is also doubtful here. You cannot use a paired test as this is not paired data.
I would here use a permutation ...
I interpret your description as follows:
For the pre-training measurements you have three sessions.
In each session you measured English minus Polish duration.
Call the three differences $X_1, X_2, X_3.$ Their mean is 63ms.
From what you say, I'm not sure of about the standard deviation
of these three differences. Maybe as large as 20, likely smaller.
My answer assumes unpaired two-sample t-tests (assuming equal variance). The settings are as follows;
The sample size, mean, and standard deviation of group A are $n_a$, $m_a$, $sd_a$, and
The sample size, mean, and standard deviation of group B are $n_b$, $m_b$, $sd_b$, respectively.
Where the standard deviation is calculated using the n-1 method.
I spent some time on this and i have to say it was a bit hard understanding the research question and data structure. So I'm shooting a bit in the dark here.
Based on this line:
"In other words, an individual is assumed to give a significantly higher score of guilt for guilt-inducing advertisement compared to the shame-score."
Shame ads are not needed, ...
I believe I misunderstood your question the first time around.
As pointed out by others, there is usually not a recipe for how or when to perform multiple comparison testing. In order to have a better understanding for what might be appropriate in your case, we would need more information about your study design and what constitutes a "test."
Suppose you ...
If the goal is to conserve the family wise error, you should consider each cell a unique test. So, if it makes a p-value it's a test. Though with over 1,000 tests, you ought to consider controlling the false discovery rate. In either case, it's still focused on each unique row and column combination.
Depending on your data you could possibly do a paired sample t-test, alternative a repeated measure ANOVA.
so it would basically measure the difference between the times they saw the green card. I don't know fore sure if it would be suitable, but its better then a regular t-test in this situation in my opinion.
Mixed models can be used to test for both between- and within-subjects effects. Both are treated in the same manner as columns of the design matrix for the fixed effects. In the former case, the value of the covariate will be the same for all the measurements of a particular subject (e.g., sex), and in the latter case the covariate will have different values ...
I'd rather err on the safe side.
By using a one-sided test, you are basically excluding any possibility of detecting that the treatment group is actually faster. Sometimes it makes sense (e.g., road safety regulations: We forbid texting only if we can prove that it makes driving less safe. We don't intend to make texting-while-driving compulsory, even if it ...