1
00:00:00,600 --> 00:00:01,520
Speaker 1: Welcome back everybody.

2
00:00:01,520 --> 00:00:06,599
Speaker 2: Today, we're going to be looking at statistical concepts. So

3
00:00:07,280 --> 00:00:11,240
effective research design psychology requires unligning the chosen methodology with

4
00:00:11,320 --> 00:00:16,559
a specific research question while acknowledging inherent limitations. Fundamental questions

5
00:00:16,600 --> 00:00:19,480
to guide design selection include what is the research question?

6
00:00:20,359 --> 00:00:24,039
Speaker 1: What is feasible? What is ethical? What is valid?

7
00:00:24,760 --> 00:00:26,600
Speaker 2: So here's some of the concepts you need to know.

8
00:00:26,879 --> 00:00:29,920
Central tendency which refers to measures that describe the typical

9
00:00:30,079 --> 00:00:33,920
representative score in a data set, typically the mean, the.

10
00:00:33,840 --> 00:00:36,000
Speaker 1: Median, or the mode.

11
00:00:36,439 --> 00:00:41,439
Speaker 2: Variability quantifies express or dispersion of scores around the central value,

12
00:00:41,759 --> 00:00:45,679
assessed through metrics such as standard deviation, varians, and range

13
00:00:46,399 --> 00:00:51,799
distributions characterized by skewness, asymmetry and critosis, which is pikedness

14
00:00:51,960 --> 00:00:55,039
with the normal distribution serving as the reference curve from any.

15
00:00:54,880 --> 00:00:56,039
Speaker 1: Infra ritual procedures.

16
00:00:56,799 --> 00:01:01,640
Speaker 2: Hypothesis testing involves comparing a null hypothesis it's an alternative hypothesis.

17
00:01:01,719 --> 00:01:05,760
Will expand on these topics later. Folks relying on p values,

18
00:01:05,920 --> 00:01:07,159
alpha levels.

19
00:01:06,799 --> 00:01:07,920
Speaker 1: And statistical power.

20
00:01:08,799 --> 00:01:12,359
Speaker 2: Errors and testing encompass type one errors, false positives and type.

21
00:01:12,159 --> 00:01:13,400
Speaker 1: Two false negatives.

22
00:01:14,000 --> 00:01:17,480
Speaker 2: Effect sizes like Cohen's d R squared or N squared

23
00:01:17,560 --> 00:01:21,359
indicates some magnitude and practical meaningfulness of a finding beyond

24
00:01:21,480 --> 00:01:27,120
mere statistic significance. Confidence intervals provide a range estimate for

25
00:01:27,159 --> 00:01:31,879
the true population parameter, offering greater insight than p values alone.

26
00:01:33,120 --> 00:01:37,159
Statistics must inform research designs, sample sized planning, and result

27
00:01:37,280 --> 00:01:40,480
interpretation within context. So let's take a look the measures

28
00:01:40,519 --> 00:01:44,159
again of central tendency. This is under the category descriptive statistics.

29
00:01:44,239 --> 00:01:48,159
The mean is the arithmetic average calculated by summing all

30
00:01:48,200 --> 00:01:50,519
scores and dividing by the number of observations.

31
00:01:51,120 --> 00:01:52,120
Speaker 1: It is sensitive to.

32
00:01:52,159 --> 00:01:56,120
Speaker 2: Extreme values or outliers, which can distort it in skew distributions.

33
00:01:57,120 --> 00:01:58,920
Speaker 1: The middle score in a data.

34
00:01:58,640 --> 00:02:01,239
Speaker 2: Sete from lowest to highest.

35
00:02:01,280 --> 00:02:02,879
Speaker 1: It is robust to outliers and.

36
00:02:02,760 --> 00:02:09,319
Speaker 2: Particularly useful with skewed data or ordinal categorical measures. Mode

37
00:02:09,520 --> 00:02:11,879
is the most frequently occurring score. It is helpful for

38
00:02:11,960 --> 00:02:16,240
categorical data and can identify multiple peaks in multimodal distributions.

39
00:02:18,599 --> 00:02:20,719
Speaker 1: Measures of variability, they're three we mentioned.

40
00:02:20,919 --> 00:02:23,680
Speaker 2: Range is the difference between the highest and lowest score.

41
00:02:24,199 --> 00:02:26,759
It is simple but quite unstable as it relies solely

42
00:02:26,800 --> 00:02:27,520
on two values.

43
00:02:29,000 --> 00:02:31,280
Speaker 1: The variance is the average square difference from the mean.

44
00:02:31,680 --> 00:02:36,400
Speaker 2: It quantifies overall dispersions.

45
00:02:38,159 --> 00:02:40,840
Speaker 1: The next is standard deviation, the square root of the variants.

46
00:02:40,879 --> 00:02:43,280
Speaker 2: It is expressed in the original units of measurement and

47
00:02:43,319 --> 00:02:44,439
indicates how far.

48
00:02:44,360 --> 00:02:46,560
Speaker 1: Scores typically deviate from the mean.

49
00:02:47,319 --> 00:02:51,080
Speaker 2: Approximately about sixty eight percent of scores fall within one

50
00:02:51,560 --> 00:02:56,000
plus or minus standard deviation and a normal distribution. Speaking

51
00:02:56,039 --> 00:02:59,680
of distributions, we'll look at two different types. The skewness

52
00:02:59,719 --> 00:03:03,599
is describing the asymmetry. A positive skew, which skews to

53
00:03:03,639 --> 00:03:07,080
the right, features a tail extending to the right, so

54
00:03:07,439 --> 00:03:09,000
things like income distributions.

55
00:03:10,000 --> 00:03:12,879
Speaker 1: A negative skew was left skew features a tail to

56
00:03:12,960 --> 00:03:13,319
the left.

57
00:03:13,439 --> 00:03:18,479
Speaker 2: Certain test scores with sealing effects cretosis. It describes peakedness

58
00:03:18,479 --> 00:03:22,039
and tail heaviness. Relative to a normal distribution. You have

59
00:03:22,120 --> 00:03:25,599
what they call a leptocurtic distribution, which are tall and skinny,

60
00:03:26,039 --> 00:03:31,840
more scores at extremes planet curtic, which are flat fewer extremes.

61
00:03:32,639 --> 00:03:35,000
The normal distribution is called mesocurtic.

62
00:03:35,000 --> 00:03:35,719
Speaker 1: If you're curious.

63
00:03:35,719 --> 00:03:39,919
Speaker 2: It serves as a benchmark from many statistical tests. You'll

64
00:03:39,960 --> 00:03:44,599
see the bell curve in the normal distribution. Inferential statistics

65
00:03:44,599 --> 00:03:49,199
and hypothesis testing. Inferential statistics allow researchers to draw conclusions

66
00:03:49,240 --> 00:03:51,159
about populations from sample data.

67
00:03:51,960 --> 00:03:54,039
Speaker 1: The null hypothesis.

68
00:03:53,280 --> 00:03:55,879
Speaker 2: That's H with a zero states there is no difference

69
00:03:56,000 --> 00:03:59,639
or no effect. Alternative hypothesis H with a one st

70
00:03:59,840 --> 00:04:02,719
there is a difference or effect. Research designs aim to

71
00:04:02,759 --> 00:04:07,319
disprove the null rather than directly proving the alternative. The

72
00:04:07,360 --> 00:04:10,319
P value is the probability of observing the obtained results

73
00:04:10,400 --> 00:04:14,479
or more extreme, assuming the null hypothesis is true. If

74
00:04:14,520 --> 00:04:17,879
the p is less than alpha, commonly points zero five,

75
00:04:18,360 --> 00:04:21,360
the null is rejected, indicating a five percent risk of

76
00:04:21,399 --> 00:04:25,399
type one error. Type one error means rejecting a true

77
00:04:25,519 --> 00:04:32,160
null hypothesis false positive, so in other words, you're saying

78
00:04:32,240 --> 00:04:36,600
the no hypothesis isn't true. The probability of observing the

79
00:04:36,639 --> 00:04:39,600
obtained results, assuming the null hypothesis true, again as the

80
00:04:39,639 --> 00:04:40,639
P value.

81
00:04:41,319 --> 00:04:45,000
Speaker 1: If p alpha p value, the null is rejected.

82
00:04:44,560 --> 00:04:46,839
Speaker 2: If it's less than point zero five, indicating a five

83
00:04:46,879 --> 00:04:50,160
percent risk of type one errors Type two errors is

84
00:04:50,199 --> 00:04:51,399
feeling to rejecting.

85
00:04:51,120 --> 00:04:54,839
Speaker 1: False null hypothesis, So false.

86
00:04:54,720 --> 00:04:59,279
Speaker 2: Negative statistical power represents the probability of detecting a true effect.

87
00:05:00,120 --> 00:05:03,439
Researchers typically aimed for power greater than point eight zero,

88
00:05:04,800 --> 00:05:10,639
which increases with larger sample sizes. Effect size quantifies the

89
00:05:10,680 --> 00:05:14,079
strength of an observe relationship or difference. One type of

90
00:05:14,079 --> 00:05:18,240
measurement is Cohen's d standard mean difference between two groups.

91
00:05:18,360 --> 00:05:21,319
Zero point two equals small, zero point five is medium,

92
00:05:21,319 --> 00:05:24,639
and zero point eight is large. R squared is coefficient

93
00:05:24,639 --> 00:05:29,319
of determination. It's looking for the meaningfulness the proportion of

94
00:05:29,399 --> 00:05:33,839
variants explained by a predictor and squared a ETA squared,

95
00:05:33,839 --> 00:05:36,800
as they call it. Ata used in a nova indicates

96
00:05:36,800 --> 00:05:41,959
proportion of variants accounted for by the independent variable. Confidence

97
00:05:42,279 --> 00:05:45,680
intervals a arranged likely to contain the true population parameter,

98
00:05:46,680 --> 00:05:50,560
So think of ninety five percent confidence interval. Narrow Narrower

99
00:05:50,600 --> 00:05:55,240
intervals reflect greater precision. Unlike P values, confidence intervals can

100
00:05:55,279 --> 00:05:57,879
be both direction and magnitude. In some say it's more

101
00:05:57,879 --> 00:06:02,079
important than the P value. Connecting statistics to research design,

102
00:06:02,160 --> 00:06:05,399
sample size planning via power analysis is essential prior to

103
00:06:05,480 --> 00:06:09,639
data collections. Smaller samples increase error and reduce power. They

104
00:06:09,680 --> 00:06:13,480
elevate the type two error risk. Design choices such as

105
00:06:13,519 --> 00:06:18,680
within subjects, have higher sensitivity, so lower type two errors

106
00:06:18,759 --> 00:06:22,560
fewer participants versus between subjects, which requires more participants but

107
00:06:22,639 --> 00:06:24,680
avoids carryover effects.

108
00:06:25,600 --> 00:06:27,399
Speaker 1: These influence statistical outcomes.

109
00:06:27,480 --> 00:06:32,319
Speaker 2: Researchers must control confounds through randomization, matching or statistical adjustment.

110
00:06:33,199 --> 00:06:39,839
When interpreting results, avoid over reliance on significance. Instead evaluate

111
00:06:39,839 --> 00:06:44,040
effect strength, contextual meaningfulness, generalizability, and study limitations.

112
00:06:44,560 --> 00:06:46,680
Speaker 1: Let me give you an example APA style reporting.

113
00:06:46,720 --> 00:06:51,240
Speaker 2: Participants in the CBT group reported significantly fewer panic symptoms.

114
00:06:52,000 --> 00:06:54,560
The mean was four point two, standard deviation is one

115
00:06:54,600 --> 00:06:58,319
point one. Then controls the mean is six point seven,

116
00:06:58,360 --> 00:07:02,160
standard deviation one point three, The t is twenty eight

117
00:07:02,240 --> 00:07:04,639
equals three to twenty three. The p value was point

118
00:07:04,800 --> 00:07:08,680
zero one, and the effects size the COENSD was zero

119
00:07:08,720 --> 00:07:12,399
point seven eight, which was what large On that end. Statistics,

120
00:07:12,439 --> 00:07:16,360
when applied thoughtfully, transform raw data into actionable insights that

121
00:07:16,439 --> 00:07:20,399
advance psychological science. Let's give you some examples of some

122
00:07:20,439 --> 00:07:22,639
of the terms we've talked about. So the mean again

123
00:07:22,720 --> 00:07:25,319
is the average. So in a study of anxiety scores

124
00:07:25,800 --> 00:07:28,079
scals zero to forty among therapy clients, a mean of

125
00:07:28,120 --> 00:07:31,439
twenty eight suggest the typical client experience is moderate to

126
00:07:31,639 --> 00:07:35,240
high anxiety, especially if it's based off of the BDI

127
00:07:36,720 --> 00:07:39,160
media and is the middle value and order data. So

128
00:07:39,240 --> 00:07:41,759
an income data for a community mental health of sample

129
00:07:42,079 --> 00:07:45,839
skewed by high earners. The media better represents central income

130
00:07:45,920 --> 00:07:49,439
than the mean. How about the mode well In a

131
00:07:49,519 --> 00:07:52,759
survey of preferred coping strategy, social support being the mode

132
00:07:52,759 --> 00:07:57,360
indicates it is the most frequently endorsed strategy. Standard deviation.

133
00:07:57,680 --> 00:07:59,879
Let's look at this. It's the average deviation from the mean.

134
00:08:00,360 --> 00:08:02,759
So you have depression scores with a mean of fifteen.

135
00:08:03,040 --> 00:08:06,519
Standard deviation of five indicates scores fall between ten and

136
00:08:06,680 --> 00:08:11,560
twenty in larger SD or standard deviation signals greater heterogeneity

137
00:08:11,639 --> 00:08:12,360
and symptoms.

138
00:08:12,360 --> 00:08:16,839
Speaker 1: Severity skewness is an example.

139
00:08:17,000 --> 00:08:20,439
Speaker 2: Like the reaction time data and cognitive psychology often shows

140
00:08:20,480 --> 00:08:27,399
positive skew due to occasional very slow responses. Critosis and examples.

141
00:08:27,439 --> 00:08:33,919
Intelligence test scores approximating normal mesocurtic distribution allow standard parametric tests.

142
00:08:35,159 --> 00:08:39,960
No hypothesis. An example would be is no effects. Or

143
00:08:40,000 --> 00:08:42,559
an example would be CBT has no effect on depression

144
00:08:42,600 --> 00:08:46,320
scores compared to weight less control alternatives.

145
00:08:46,360 --> 00:08:47,200
Speaker 1: There is an effect.

146
00:08:47,360 --> 00:08:50,879
Speaker 2: CBT reduces depression scores relative.

147
00:08:50,480 --> 00:08:51,399
Speaker 1: The control type.

148
00:08:51,440 --> 00:08:54,080
Speaker 2: One is a false positive, concluding a new therapy is

149
00:08:54,120 --> 00:08:59,320
effective when it's not potentially leading to widespread adoptions. So

150
00:08:59,399 --> 00:09:04,120
basically saying that the Noull hypothesis has been rejected when

151
00:09:04,120 --> 00:09:08,159
it hasn't. Type two error is false negative failing to

152
00:09:08,200 --> 00:09:11,799
detect it be beneficial effective mindfulness on stress missing an

153
00:09:11,840 --> 00:09:17,799
opportunity for clinical application. So this is saying that Noll

154
00:09:17,879 --> 00:09:23,879
hypothesis is true when it's not. Cohen's D is standardized

155
00:09:23,919 --> 00:09:28,039
effects size point eight five for exposure therapy on PTSD

156
00:09:28,120 --> 00:09:33,240
indicates a large meaningful improvement. Confidence interval range for population parameter.

157
00:09:33,320 --> 00:09:35,960
For instance, a ninety five percent confidence level for mean

158
00:09:36,000 --> 00:09:39,080
difference and treatment outcomes two point one and five point

159
00:09:39,159 --> 00:09:42,799
three suggest the true effect is positive and reasonably precise.

160
00:09:43,919 --> 00:09:45,320
Speaker 1: Here's a practice question for you.

161
00:09:45,320 --> 00:09:49,039
Speaker 2: A researcher finds that a new intervention yields point zero three,

162
00:09:49,840 --> 00:09:52,600
a P value of point zero three, which is statistically

163
00:09:52,600 --> 00:09:55,320
significant with the Coens D of zero point twenty five.

164
00:09:56,679 --> 00:10:04,320
So what is the size of the Cohen's D? If

165
00:10:04,360 --> 00:10:10,559
you said small, you're right. I hopefully this will help

166
00:10:10,559 --> 00:10:12,399
you with today's podcast. If we talked to some course

167
00:10:12,519 --> 00:10:16,720
statistical concepts, we didn't go into too much detail, too severe,

168
00:10:16,960 --> 00:10:18,440
but hopefully you learned a little bit from there.