live_tv
Livestream Starting Soon
00
Hours
:
00
Minutes
:
00
Seconds
Up next in 10
JASP vs Jamovi vs PSPP: which free stats program is best for three common tests?
Dec 4, 2025
We do a t-test, some correlations, and multiple regression in three free statistics software packages so you don’t have to try all three! See what each does best... and how easy or hard it is to work with the data!
Show More Show Less View Video Transcript
0:00
Hi, it's me, David Zatz, the guy who
0:02
runs the macstats.org,
0:05
Mac statistics software website. I'm
0:08
here to guide you through which one is
0:11
better for us. JASP, Jamovi or PppP and
0:17
I'm just going to focus in this case on
0:19
some things that I personally have done
0:21
quite commonly. TA tests, correlations,
0:24
and regressions. So, without further
0:26
ado, I'm going to turn it over to
0:28
microphone me as opposed to video me.
0:31
So, here we are in PSP first. And this
0:35
is an open-source uh replacement for an
0:39
older version of SPSS. And this is a
0:42
database of houses in a small town,
0:44
which I found that was publicly
0:46
available. and it has the sales price,
0:49
the listing price, discount days, number
0:52
of beds, number of bathrooms, type of
0:54
house, garage. So, let's try analyze
0:58
compare means independent samples t
1:01
test. And we'll use number of bedrooms
1:05
first. And this lets us decide that
1:08
we're going to compare the two-bedroom
1:10
houses to the three-bedroom houses.
1:12
Actually, let's make it fourbedroom
1:14
houses to threebedroom houses. So we can
1:17
use almost any variable as a group in
1:20
tests. We don't have to just use ones
1:23
with two values.
1:26
And wow, was that fast instant much
1:29
faster than SPSS itself. Now here we
1:32
have the same thing that we have roughly
1:34
in SPSS. But they both give you the
1:36
averages for the two group. Uh and then
1:39
to find out is that difference
1:41
significant? There's two types of t
1:44
tests. There's two formulas. There's the
1:46
original student formula and there's
1:48
Welch's. And if you look, equal variance
1:51
is assumed, equal variance is not
1:53
assumed is the difference between the
1:54
two. And what you're supposed to do is
1:56
look at this number here, the
1:58
significance of Lavine's test to see if
2:01
the variances are equal. If it's
2:03
significant, you cannot assume that
2:05
they're equal. And you have to use this
2:07
second level here to get to this
2:09
significance, which tells you that it is
2:12
definitely not equal. Now, obviously, if
2:15
you've been doing this long enough, you
2:18
can just look here, and if they're both
2:19
clearly significant, it doesn't really
2:21
matter which one you look at. In this
2:23
case, they're both the same. It will
2:25
give you the results of both formulas,
2:28
and you have to pick the one that is
2:30
relevant. Now, you can ask, why doesn't
2:32
it do it for you? And I honestly don't
2:35
know. SPSS doesn't do it for you either,
2:38
but it can get even crazier. Let's take
2:40
a look. Let's take a look and see how
2:42
Jamovi does it now. So t tests
2:45
independent samples
2:48
sold price uh for grouping variable on
2:51
this one. I'm using whether or not you
2:53
have a garage and attach garage. And
2:56
look here. This is what happens if you
2:59
do bedrooms. It it gives you the same
3:01
error message. You have to have exactly
3:04
two levels. So you have to clean up your
3:06
data before you can do this. You can't
3:08
just pick two uh numbers out of the
3:12
group and use those, which is a bit of a
3:14
nuisance. But let's go here to data. And
3:17
this is pretty easy to do. Um we add
3:20
nine as a missing value.
3:23
And now you see that the nines are
3:25
graded out are grayed out. So let's go
3:27
back to the analysis and we'll go back
3:31
to test.
3:33
And we'll go back to put sales price in
3:35
here. and uh garage binary. I already
3:39
created this variable just for this
3:41
demonstration. And here we also want to
3:43
put in the descriptives
3:47
because you want to see the means. So
3:49
here we see that the houses they had 78
3:52
sold without attached garages, 100 with
3:55
attached garages. The average price for
3:58
the houses with attached garages is much
4:00
higher. It's about 70
4:04
uh let's say $73,000 more than the ones
4:08
with detached or no garage. And if you
4:11
look for significance here, it is uh P
4:14
is 0.011. So if you're just doing this
4:17
one statistic and you've got your P at
4:20
0.05, you are free and clear. Here,
4:22
here's one problem though. This is easy
4:25
to overlook. Take a look down here in
4:27
the fine print. Lavine's test is
4:29
significant. So there is a violation of
4:32
the assumption of equal variances. So
4:34
the two groups both have different
4:35
variances which means that this P of
4:38
0.011 completely irrelevant. You can't
4:41
use it. You have to go down here and
4:43
select Welch's T and use that. So you uh
4:47
Welch's P sorry. And here you see it's a
4:50
better level of P for you. It's more
4:52
significant. So, it's very important
4:55
when you're using Jamovi or JASP for
4:57
that matter to look for these little
4:59
footnotes and then to run the second
5:01
test. Um, the question is, what if we
5:04
were to run this with both of these
5:06
selected? If we just did all of them
5:07
that way, we'll start over fresh.
5:11
We want students and Welches. We'll go
5:14
back to sold and we'll go back to garage
5:17
binary.
5:20
So, here you've got both. And it tells
5:22
you with that little note not to use the
5:25
Welch's test. All right, let's take a
5:28
look at JASP. So t tests, you have to
5:33
remember to stay in the classical area
5:34
now unless you want to do basian
5:36
analyses. So let's try doing both here.
5:39
Student and Welsh
5:43
and we do sold garage binary.
5:47
Uh here again I have to set the missing
5:49
value. I thought I had already done
5:51
that. So, garage binary missing values
5:55
use custom missing values
5:59
make nine missing. All right, go back to
6:01
analyses and again we have to say
6:04
descriptives. I don't know why that's
6:05
not standard. I can't imagine running
6:07
tea tests and not wanting to know the
6:10
two averages and also not wanting to
6:13
know the two ends because you really do
6:15
need to know how many people are in each
6:17
group to make sure that you've got
6:19
enough. Uh the numbers big surprise are
6:22
exactly the same. But look what JASP
6:24
misses. JASP does not tell you which of
6:28
these two formulas to use.
6:31
If I shut off Welch, it just tells me
6:34
students, but it doesn't tell me that
6:36
there's a problem. You have to also look
6:38
for assumption check for equality of
6:40
variances.
6:42
Uh we'll use Lavines to be consistent.
6:45
And then you look at this and then you
6:47
pick a number. So it's a lot like SPSS
6:49
and PSP in that regard except it really
6:52
should be coming up automatically. And
6:55
when you think about it, it would make
6:56
sense for it to just run it. If students
6:59
test is inappropriate, put in Welch's
7:02
test and vice versa and tell you about
7:05
it when it's done. Okay. Now, we're
7:07
going to do correlations. Uh we'll start
7:09
out this time with Jimovi. We'll do it
7:11
in reverse order. Say correlations.
7:14
Uh we'll pick the date sold, the listed
7:16
price, the discount, number of days it
7:19
was sold, beds and baths. Put them all
7:22
into variables. Not going to partial
7:23
anything out this time. And it gives you
7:26
this interesting format which is kind of
7:28
unusual. It's really a more efficient
7:31
format, but one which is in some ways a
7:33
little tougher to look at. You can make
7:35
it flag significant correlations.
7:38
Really, what I'd also like to see is the
7:41
number of people in each of these.
7:44
And there it is. Click on sample size
7:46
and it comes up. I I'm again a little
7:48
surprised that's not the default. I I
7:50
generally want to see this. Uh now if
7:53
you want to see the more traditional
7:54
look, you just take away display
7:57
pair-wise and here's your traditional
7:59
table. The N would be nice to see
8:01
because then you'd know how much how
8:03
many of these you have the data for. But
8:06
this this is how I'm used to seeing it.
8:07
So maybe I just think it's better
8:09
because I'm used to it. Let's take a
8:11
look now at Jamovi
8:13
correlation matrix. And I don't really
8:16
use Jamovi very much. So, I'm doing all
8:18
of this as I go along.
8:22
And that's something just to keep in
8:23
mind is that these programs are very
8:25
easy to learn how to use without a lot
8:28
of trouble. This is exactly how I'm used
8:30
to seeing it except that I'm used to uh
8:33
again having the number of people here
8:37
and I'm more used to seeing end and
8:39
degrees of freedom, which is kind of an
8:41
interesting way to do it. Uh you can
8:43
shut off significance if you want it to
8:45
be really tight. You can flag things
8:46
that are significant and finally PSP
8:52
analyze
8:54
by variate correlations.
8:57
Uh let's see we don't want address
8:59
obviously sold.
9:02
You can also flag significant ones
9:04
again. And this comes up in the way that
9:07
I uh in the SPSS fashion that I'm used
9:10
to seeing it. It's on both sides. So
9:14
you're seeing every number is reported
9:15
twice. So listed versus price is the
9:19
same as price versus listed. Now one
9:22
thing I do want to mention on these is
9:23
let's suppose that you're picking your
9:25
variables. If you have a much larger
9:28
data set and sometimes I do work with uh
9:30
data sets that have like 100 different
9:32
variables in them or more. Uh you will
9:35
want to make it easy to find different
9:38
variables. So, one way that you can do
9:40
that is by having it so that when you
9:41
press the letter G, it goes to garage
9:45
and so on. And then as you type in, you
9:48
can get more. So, you can see that if I
9:50
look for beds, uh, let's go back up
9:52
here. Let's say I'm looking for
9:53
bathrooms. I type B and it shows bed. I
9:55
type BA and I get baths. So, a lot more
9:59
convenient than it would be if it didn't
10:01
do that. Uh, so let's take a look at
10:04
Jimovi and we'll put all these guys back
10:06
and I'll try the same thing. I look for
10:09
beds. Oh, look, nothing's happening. So,
10:12
let's try JASP. And this is one of the
10:15
reasons I really use JASP more often
10:18
now.
10:21
So, you see it's the same rule in JASP.
10:24
I type ba, it takes me right to baths.
10:27
That might not seem like a big deal, but
10:29
when you've got a 100 variables or 200
10:31
variables, it is really hard to find
10:33
stuff otherwise. All right, our final
10:35
one is regression. We're going to do
10:38
linear regression, and we'll start out
10:40
with JASP. Now, one fun thing about
10:43
JASP, let's say that you want to uh see
10:46
how many days a house is on the market.
10:48
What what is the reason for that? That's
10:50
days.
10:52
And you want to see whether the sales
10:54
price,
10:56
the listed price, the discount, actually
10:59
there's no point putting a sales price
11:00
in. Whether the listed price, the
11:02
discounted price, or the number of beds
11:05
and baths is most important.
11:08
You can either enter them all at once.
11:10
You can or you can put them in backward,
11:12
forward, or stepwise. Stepwise tests it
11:15
every which way and sees which explains
11:19
the most variance. So you can explain
11:22
30.6% of the variance in days using the
11:27
discount and the listed price. That's
11:30
it. That that's pretty impressive. Now
11:32
this is very good for exploratory data.
11:34
You don't want to use it too much. You
11:35
don't want to abuse it. Let's say that I
11:37
wanted to make sure that uh the listed
11:40
price loads in first. If we want to see
11:43
what happens when listed price comes
11:45
goes in first instead of discount, we go
11:48
to enter instead and then we set up
11:50
different models and the first model has
11:54
listed price and the second one has
11:56
both. And you can see that the uh end R
11:59
squar is the same but the first R squar
12:03
has gone from 28% to just 10%. So, the
12:08
listed amount, the the asking price does
12:11
not have a huge impact on how many days
12:14
it sits on the market, but the discount,
12:16
how much they're willing to come down
12:18
does. So, let's go back and I'll just
12:21
say uh stepwise again so that you can
12:23
see it.
12:25
Let's go into Jamovi uh regression,
12:28
linear regression.
12:30
Uh we want to know the days on the
12:32
market. We've got the discount and the
12:34
listed price.
12:37
And let's see. Uh-oh, there's no
12:39
step-wise. It turns out that there is no
12:42
way to actually do step-wise regression.
12:44
You only have enter. So, you have to do
12:47
it, you can do it both ways here, listed
12:49
and then discount, and you get the same
12:51
numbers as in Jamovi.
12:54
Uh, or you can change the blocks around
12:56
and do it the other way. So, it's
12:59
discount and then listed. You can't just
13:02
have it go through and decide, do
13:05
bedrooms or bathrooms matter? You can
13:07
put them all in and then have more
13:10
blocks.
13:13
Let's add new block.
13:18
See,
13:25
they don't make it easy.
13:31
There we go. And it'll run it all, but
13:35
it won't the last two won't be
13:37
significant. So, how about PSP,
13:41
the last one?
13:46
Uh, let's see. Days on the market and
13:49
then
13:51
listed in discount here. And yep, again,
13:54
there's no step-wise regression. It's
13:56
just going to be enter. And what's more,
13:59
it appears that it can't even let you
14:02
you can't even enter in blocks. You have
14:04
to enter them both at once, which means
14:07
that the regression is nowhere near as
14:09
good as it should be here. There might
14:11
be if you use syntax, but I don't
14:13
believe that there is. I really like the
14:15
fact that PPP has a very easy way to
14:19
recode any of the variables. It's much
14:21
easier to deal with in terms of
14:24
manipulating the variables themselves.
14:26
It doesn't make you create a new
14:27
variable just to do a t test. I like the
14:30
fact that JASP lets you do stepwise
14:33
regression. And I like the general look
14:35
and feel of Jamovi. I like the fact that
14:37
to select a variable in PPP and JASP,
14:40
you just type in a couple of the
14:42
letters. Um I like the fact that in uh
14:45
Jamovi, if you run both Welches and
14:49
students tests, uh it will tell you
14:52
which one to use. if you look really
14:55
hard. Well, that's it for the summary.
14:57
You can read more at maxstats.org, which
15:00
I will keep updated because I can't
15:02
update these videos, but I can update
15:05
the website. Talk to you later.
#Science