I have done t-test (two-sample assuming unequal variances). In this case, I am rejecting the null hypothesis. There is a difference between the groups. I am just wondering if there is any way to find which group is performing better?
$\begingroup$
$\endgroup$
5
-
1$\begingroup$ At the moment, the title and text don't ask the same question. $\endgroup$– Sal MangiaficoFeb 26, 2020 at 16:35
-
3$\begingroup$ "Normal hypothesis is true" means you don't expect either group to perform better. But when you do conclude one group performs better, there is an insanely simple way to figure out which one it was: compare their means. $\endgroup$– whuber ♦Feb 26, 2020 at 16:43
-
$\begingroup$ @SalMangiafico I am really sorry! Thanks for pointing that out. I have rectified it now. $\endgroup$– Ashish KumarFeb 26, 2020 at 18:14
-
3$\begingroup$ For one, is low bad or good? Overall or in the population? Reducing salt intake is generally good, unless you're hyponatremic. $\endgroup$– AdamOFeb 26, 2020 at 18:26
-
$\begingroup$ It is total transactions done for the month of Jan vs Feb. So more, the better. $\endgroup$– Ashish KumarFeb 26, 2020 at 20:20
Add a comment
|
3 Answers
$\begingroup$
$\endgroup$
It depends on how you've set the test up.
If your test statistic has $\bar{x}_1 - \bar{x}_2$ in the numerator, then your test statistic is looking at the mean of group 1 minus the mean of group 2. A negative difference means group 2 has a higher mean.
So to know which group did better, look at the sign of the test statistic and the order in which the group means were used.
answered Feb 26, 2020 at 18:21
$\begingroup$
$\endgroup$
12
It helps to check the means and standard deviations of your groups as well as to visualize the differences. I really like the Datanovia approach to things, where you also just add test data onto plots. An example is given below using R. First we can load the required libraries
#### Load Libraries ####
library(datarium)
library(tidyverse)
library(rstatix)
#### Check Mean/SD of Groups ####
genderweight %>%
group_by(group) %>%
summarise(mean.weight = mean(weight),
sd.weight = sd(weight))
giving us these summary statistics
# A tibble: 2 × 3
group mean.weight sd.weight
<fct> <dbl> <dbl>
1 F 63.5 2.03
2 M 85.8 4.35
We can see from the descriptive data here that the female group has a lower weight on average and varies less than the male group, so if we have a significant t-test, we should expect that it is because the female group is lower on average than the male group. We can run a t-test to check quickly:
#### Run T-Test ####
t <- genderweight %>%
t_test(formula = weight ~ group)
And indeed this is true...
# A tibble: 1 × 8
.y. group1 group2 n1 n2 statistic df p
* <chr> <chr> <chr> <int> <int> <dbl> <dbl> <dbl>
1 weight F M 20 20 -20.8 26.9 4.30e-18
The negative sign indicates that the female group (the reference group here) has a lower mean than the male group. We can look at this difference directly with a boxplot and even add our t-test data to contextualize these differences:
#### Visualize Differences ####
genderweight %>%
ggplot(aes(x=group,
y=weight,
fill=group))+
geom_boxplot()+
labs(title = "Weight By Gender",
subtitle = get_test_label(t, detailed = T))
And now you can see the which group has a smaller and which a larger mean and that the female group has a much narrower box because of its lack of variation compared with males:
Edit
Nick brings up a very compelling point about boxplots in the comments. While this gives some perspective on what the average weights are with each group, they are more based around the interquartile range and medians of the data, which are not directly comparable to means and standard deviations (especially in the case of skewed data). Some other visualizations may serve this purpose better (of which he mentions quantile plots with overlayed means), but at the minimum, inspecting the means and SDs as well as giving yourself graphical representations of the data help.
answered May 25 at 6:54
-
$\begingroup$ This is helpful but you're following a multitude here -- all of whom should know better -- to give a box plot as context for a comparison of means without either showing means added to the plot or flagging that your box plot does not show means directly. It's obvious to all experienced people from the box plots alone that the means are different -- and (I guess) will be obvious to many learners too. And naturally you've calculated the means and they can be compared with the graph. Nevertheless this exemplifies a common indirectness. $\endgroup$– Nick CoxMay 25 at 7:16
-
1$\begingroup$ The best kind of plot for comparison of means will show those means in a data-rich context. We can have discussions of what works best, which will pivot on personal taste and tribal habits, and box plots with added means would be one candidate. My personal favourite of quantile plots with added means is evidently a minority taste common in no tribe. Box plots have swung from being helpful extras to being used routinely when often other methods are more informative and no less effective. $\endgroup$– Nick CoxMay 25 at 7:18
-
1$\begingroup$ I go hot and cold on the use of densities, meaning specifically kernel density estimates. Densities are great if you understand something about how they are calculated and about their limitations. But I am repeatedly shocked by poor practices, for example with so-called violin plots used naively. It seems common that people use defaults without realising that (a) the defaults are often not smart (b) there are other choices (c) defaults often smear probability into impossible regions (d) defaults are often poor choices for variables that are bounded (as (c)) and/or very skewed. (ctd) $\endgroup$– Nick CoxMay 25 at 7:57
-
1$\begingroup$ stats.stackexchange.com/questions/205629/… is one example. I've posted about quantile plots here several times and re-discovering the thread most pertinent here is hard. $\endgroup$– Nick CoxMay 25 at 8:06
-
1$\begingroup$ I am not a statistician and so familiar with the struggles of non-statisticians to understand and use statistics. Professionals as well as students aren't always comfortable even with the idea that you need to think about bin width and origin for histograms, and the difficulties aren't less with kernel density estimation. A common social difficulty with box plots is that statistical people often don't explain what is shown by a box plot, as they regard them as obvious. This is particularly acute because there are so many variants on rules for which points are shown separately. $\endgroup$– Nick CoxMay 25 at 8:12
$\begingroup$
$\endgroup$
1
The group that has the mean that is in the "better" direction did better on average. The t-test lets you assess the strength of evidence against the null hypothesis, and so if you got a small enough p-value then it is reasonable (in some sort of proportion to the smallness of the p-value) to think that one group did better.
Seeing which group did better on average needs no statistics beyond the mean. Just graph the data. Never make inferences on the basis of data without examining the data in some sort of visual display.
-
$\begingroup$ +1 for the simple answer, though I would contend that it is important to have other additional descriptive statistics to contextualize just how precise this mean is (as highlighted in my answer and certainly not limited to what I said there). $\endgroup$ May 25 at 6:57