1
00:00:00,600 --> 00:00:01,520
Speaker 2: Welcome back everybody.

2
00:00:01,560 --> 00:00:06,519
Speaker 1: Today we're going to be looking at assessments, research methods,

3
00:00:06,519 --> 00:00:09,759
and statistics for today's or this week's study guide. Will

4
00:00:09,759 --> 00:00:12,080
create questions for you towards the end of the week

5
00:00:12,640 --> 00:00:15,640
to review it. Hopefully that you like the new style.

6
00:00:16,359 --> 00:00:18,519
We go over a topic and then during the middle

7
00:00:18,559 --> 00:00:21,839
of the week we will release ten to fifteen practice

8
00:00:21,920 --> 00:00:25,600
questions to kind of review it. So let's look at

9
00:00:25,600 --> 00:00:29,480
statistics and research methods, so research designs, that's the framework

10
00:00:29,519 --> 00:00:31,640
of a study. You must be able to distinguish between

11
00:00:31,679 --> 00:00:37,640
an experimental study, quasi experimental, and correlational designs, specifically focusing

12
00:00:37,679 --> 00:00:42,520
on their ability to establish causality versus mere relationships. Also,

13
00:00:42,600 --> 00:00:45,320
you need to know statistical significance p values and measure

14
00:00:45,359 --> 00:00:50,439
the probability that an observed difference occurred by chance limitations

15
00:00:50,439 --> 00:00:52,600
and statistically a significant result.

16
00:00:53,960 --> 00:00:56,719
Speaker 2: Point zero five or less than point zero five.

17
00:00:56,799 --> 00:00:59,320
Speaker 1: It does not mean the finding is practically important in

18
00:00:59,320 --> 00:01:03,000
a clinical sety what it does. What does say that

19
00:01:03,119 --> 00:01:07,040
is the effect size, sometimes called the clinical significance, so.

20
00:01:07,120 --> 00:01:10,200
Speaker 2: Standardized measure the magnitude or strength of an effect.

21
00:01:10,799 --> 00:01:15,280
Speaker 1: This tells you the protocol significance of a treatment, validity,

22
00:01:15,359 --> 00:01:16,799
the accuracy.

23
00:01:16,239 --> 00:01:18,280
Speaker 2: And truthfulness of a study's conclusions.

24
00:01:19,359 --> 00:01:23,239
Speaker 1: Threats to validity or methodological flaws that compromise your findings.

25
00:01:23,760 --> 00:01:29,959
Another key triple P threat includes history, external events, altering outcomes, maturation,

26
00:01:30,239 --> 00:01:34,760
natural changes and participants over a course of time, selection bias.

27
00:01:34,799 --> 00:01:35,400
Speaker 2: Something else that.

28
00:01:35,400 --> 00:01:38,719
Speaker 1: Can affect the threats to validity is systematic differences between

29
00:01:38,799 --> 00:01:45,200
groups at the start. Ethical research practices safeguards protecting participants.

30
00:01:45,760 --> 00:01:50,799
Essential components include informed consent, confidentiality protections, minimizing risk, and

31
00:01:50,879 --> 00:01:57,280
the regulatory oversight of IRBs. Statistical tests mathematical tools used

32
00:01:57,280 --> 00:01:59,799
to analyze data. You must known't when to apply specific

33
00:01:59,799 --> 00:02:03,359
ten tests based on your variables. A T tests or

34
00:02:03,359 --> 00:02:06,280
in a NOVA is used when comparing group means, comparing

35
00:02:06,319 --> 00:02:11,120
treatment versus control. Correlation and regression use when looking for

36
00:02:11,199 --> 00:02:17,000
continuous relationships or predicting outcomes. KI square used with analyzing

37
00:02:17,039 --> 00:02:20,159
categorical data counting the number of patients who drop out

38
00:02:20,319 --> 00:02:22,759
versus finish therapy.

39
00:02:24,000 --> 00:02:26,039
Speaker 2: Descriptive statistics.

40
00:02:27,560 --> 00:02:31,439
Speaker 1: Numerical techniques used to summarize, organize, and describe the basic

41
00:02:31,479 --> 00:02:33,400
features of a data set, so you'd be looking at

42
00:02:33,439 --> 00:02:37,879
things like the mean median, the mode, standard deviations Z

43
00:02:38,000 --> 00:02:44,639
scores things like that inferential statistics techniques that allow researchers

44
00:02:44,680 --> 00:02:49,400
to use sample data to make generalizations, estimates, or predictions

45
00:02:49,400 --> 00:02:50,639
about a larger population.

46
00:02:52,719 --> 00:02:54,199
Speaker 2: Measurement scores.

47
00:02:55,560 --> 00:02:58,960
Speaker 1: The classification system for how variables are categorized and measured.

48
00:02:59,199 --> 00:03:01,599
Speaker 2: Four types nominal, which.

49
00:03:01,400 --> 00:03:05,680
Speaker 1: Include things like categories such as diagnosis. For instance, Ordinal

50
00:03:05,840 --> 00:03:09,439
is ordered by ranks, so symptoms severity like mild, moderate,

51
00:03:09,639 --> 00:03:13,840
or severe intervals ordered categories with equal distance so there's

52
00:03:13,879 --> 00:03:19,719
no true zero, like IQ and ratio, which equal distance

53
00:03:19,759 --> 00:03:24,840
with a true absolute zero reaction time and weight number

54
00:03:24,840 --> 00:03:27,560
of Another one is research design comparisons, so if you're

55
00:03:27,560 --> 00:03:31,919
looking at true experimental design. The gold standard for establishing

56
00:03:32,000 --> 00:03:34,479
causal relationships the randomized control trial.

57
00:03:34,879 --> 00:03:36,919
Speaker 2: It requires three core features.

58
00:03:36,639 --> 00:03:40,960
Speaker 1: Random assignment, manipulation of an independent variable, and a control

59
00:03:41,000 --> 00:03:45,439
group or condition to minimize extraneous variables. The strengths to

60
00:03:45,520 --> 00:03:49,639
have high internal validity. Man validity is how much is

61
00:03:49,639 --> 00:03:51,439
a measure what it's supposed to measure, and you have

62
00:03:51,479 --> 00:03:56,000
a high internal validity effectively controls for confounding variables other

63
00:03:56,080 --> 00:04:02,159
factors that can cause problems. Limitations can reduce external validity

64
00:04:02,240 --> 00:04:02,560
due to.

65
00:04:02,560 --> 00:04:04,680
Speaker 2: High art official laboratory settings.

66
00:04:05,800 --> 00:04:09,560
Speaker 1: In other words, how does this apply outside the generalizability

67
00:04:10,000 --> 00:04:15,319
is compromised on racts or these random assignment or the

68
00:04:15,360 --> 00:04:20,680
true experimental design. It also faces strict ethical and practical constraints.

69
00:04:20,680 --> 00:04:23,920
You cannot randomly assign someone to experience trauma or abuse.

70
00:04:24,439 --> 00:04:26,800
Let me give you an example, testing whether a new

71
00:04:26,839 --> 00:04:31,639
CBT protocol reduces panic symptoms by randomly assigning fifty padic

72
00:04:31,680 --> 00:04:35,040
disorder patients to either the CB two group or a

73
00:04:35,079 --> 00:04:38,959
waitless control group and then measuring their panic levels but

74
00:04:39,160 --> 00:04:42,959
not work definition of research design that seeks to infer

75
00:04:42,959 --> 00:04:46,959
a causal relationship but lacks the crucial feature of random assignment.

76
00:04:48,480 --> 00:04:52,959
This is a quasi experimental design. Strengths more practical and ethical.

77
00:04:52,600 --> 00:04:54,319
Speaker 2: For real world clinical.

78
00:04:53,839 --> 00:04:58,199
Speaker 1: And educational settings, so higher external validity than true experiments

79
00:04:58,199 --> 00:05:01,399
because it allows you again to measure what's going on.

80
00:05:02,040 --> 00:05:07,920
Limitations lower internal validity due to selection bias, requires more

81
00:05:08,000 --> 00:05:14,160
rigorous statistical control to rule out confounds. So with quasis,

82
00:05:14,360 --> 00:05:16,360
some of the important parts of it again is that

83
00:05:16,519 --> 00:05:18,879
it seeks to infer a causal relation but lacks the

84
00:05:18,920 --> 00:05:21,759
crucial feature of random assignment. That's it's going to hurt

85
00:05:21,800 --> 00:05:25,120
a little bit the internal validity, and the external validity

86
00:05:25,160 --> 00:05:27,959
is a little bit better because you're using it in a

87
00:05:28,000 --> 00:05:33,959
more practical settings than true experiments. The text highlights three

88
00:05:34,000 --> 00:05:40,399
specific quasi experimental variations. Non equivalent control group design. Comparing

89
00:05:40,399 --> 00:05:43,519
two pre existing non randomized groups where one gets a

90
00:05:43,519 --> 00:05:44,720
treatment and the other doesn't.

91
00:05:45,439 --> 00:05:46,319
Speaker 2: You've got pre test.

92
00:05:46,360 --> 00:05:49,199
Speaker 1: Post test measures are used to catch baseline differences in

93
00:05:49,279 --> 00:05:53,879
non equivalent control group designs. An example is comparing two

94
00:05:53,959 --> 00:05:56,879
different local schools, one that naturally decides to adopt a

95
00:05:56,959 --> 00:05:59,480
mindfulness curriculum and wine that does not to see the

96
00:05:59,519 --> 00:06:06,240
impact on student anxiety. Time series design taking multiple observations

97
00:06:06,319 --> 00:06:09,759
or measurements before and after and interventions introduced. It tracks

98
00:06:09,800 --> 00:06:12,120
trends over time and is highly useful in policy or

99
00:06:12,160 --> 00:06:21,800
organizational settings. In addition to that good blost my train

100
00:06:21,839 --> 00:06:25,040
of thought there, sorry, folks. Tracking emergency as an example,

101
00:06:25,120 --> 00:06:28,199
tracking emergency room psychiatric emissions every month for two years

102
00:06:28,240 --> 00:06:32,040
before and two years after a community wide violence reduction

103
00:06:32,120 --> 00:06:37,279
initiative is launched. Number three is regression discontinuity design. It's

104
00:06:37,319 --> 00:06:39,800
a design where participants are assigned to a treatment or

105
00:06:39,839 --> 00:06:44,000
control condition based entirely on whether this score falls above

106
00:06:44,040 --> 00:06:47,839
or below a predetermined cutoff score. It allows for strong

107
00:06:47,920 --> 00:06:52,000
causal inference if its strict mathematical assumptions are met. An

108
00:06:52,000 --> 00:06:55,040
example would be testing the impact of an intensive academic

109
00:06:55,399 --> 00:06:58,959
intervention where only students who scored below a seventy on

110
00:06:59,000 --> 00:07:03,000
a standardized receive the tutoring, with those scoring above attack

111
00:07:03,279 --> 00:07:07,319
the control group. Let's take a look at non experimental

112
00:07:07,360 --> 00:07:10,680
designs and see what we come up with. On this

113
00:07:10,720 --> 00:07:16,959
side non experimental designs, one of them is it's called

114
00:07:17,000 --> 00:07:20,319
it's defined by observational research, where the investigator does not

115
00:07:20,439 --> 00:07:25,839
intervene or manipulate variables. These are excellent for identifying relationships, trends,

116
00:07:25,879 --> 00:07:29,720
or rare cases, but cannot prove causation. The text outlines

117
00:07:29,839 --> 00:07:36,199
for major non experimental types. So again, observational research includes

118
00:07:36,279 --> 00:07:40,240
correlational studies, but not all observational research is coreational studies,

119
00:07:40,240 --> 00:07:45,160
but all correlational studies are observational research. Speaking of correlational

120
00:07:45,160 --> 00:07:48,839
studies designed to measure the relationship between variables without any manipulation,

121
00:07:49,800 --> 00:07:52,639
the core metric is that pearsons are to quantify the

122
00:07:52,680 --> 00:07:55,959
strength and direction of the relationship. A plus one point

123
00:07:56,040 --> 00:07:59,439
zero is a perfect both heading in the same direction

124
00:07:59,639 --> 00:08:02,439
and verse relationship is a negative one point zero is

125
00:08:02,480 --> 00:08:08,439
perfect limitations. Again, correlation is not equal causation. You also

126
00:08:08,480 --> 00:08:09,959
have an issue with directionality.

127
00:08:10,000 --> 00:08:12,240
Speaker 2: Does X cause Y? Or does Y cause x?

128
00:08:12,759 --> 00:08:15,240
Speaker 1: And then you have third variable problems does X cause

129
00:08:15,279 --> 00:08:15,920
both X and.

130
00:08:16,000 --> 00:08:18,759
Speaker 2: Y, which can easily mislead researchers.

131
00:08:19,720 --> 00:08:22,839
Speaker 1: An example well be finding a strong positive correlation between

132
00:08:22,879 --> 00:08:26,720
a patient's sleep quality and their weekly academic exam performance.

133
00:08:27,720 --> 00:08:34,960
Is an example of a correlational study. Case studies and

134
00:08:35,000 --> 00:08:37,759
an in depth exploration of one individual group of event

135
00:08:37,960 --> 00:08:40,639
a trade off. It provides rich detail but suffers from

136
00:08:40,799 --> 00:08:46,399
very low generalizability. Example, extensively studying document and documenting memory

137
00:08:46,440 --> 00:08:49,399
and behavioral changes in a single patient who sustained localized

138
00:08:49,440 --> 00:08:55,320
hippocampal damage. Naturalistic observation the practice of observing behavior and

139
00:08:55,440 --> 00:09:00,039
natural settings without interference or manipulation by the researcher. So

140
00:09:00,039 --> 00:09:03,120
it's to the utility, highly useful in developmental animal across

141
00:09:03,120 --> 00:09:07,000
cultural research. A trade off, very hard to control variables,

142
00:09:07,000 --> 00:09:10,720
but both incredibly high ecological validity, or other words, real

143
00:09:10,799 --> 00:09:16,399
world truthfulness. Survey research is the method they use is

144
00:09:16,879 --> 00:09:20,399
self report to gather data on beliefs behaviors and attitudes

145
00:09:21,039 --> 00:09:23,600
a trade off. It's highly efficient for large samples, but

146
00:09:23,679 --> 00:09:27,000
uniquely vulnerable to bias in accuracy and low response rates.

147
00:09:28,440 --> 00:09:31,840
The design factors elements like question wording, order effects, and

148
00:09:31,879 --> 00:09:35,919
sampling strategy heavily impact the overall quality and validity of

149
00:09:35,960 --> 00:09:41,600
the data. We have advanced design structures cross sectional versus longitudinal.

150
00:09:41,799 --> 00:09:44,679
A cross sectional really common in the world of neuroscience.

151
00:09:45,159 --> 00:09:48,000
The study design it takes a snapshot of multiple groups

152
00:09:48,039 --> 00:09:51,080
at one point in time. It is fast and cost effective,

153
00:09:51,120 --> 00:09:56,320
but cannot track individual developmental trajectories and causation is also problematic.

154
00:09:57,639 --> 00:09:59,600
You see a lot of neuroscience where they take imaging

155
00:09:59,639 --> 00:10:01,320
on the brain and they say, oh, look, people with

156
00:10:01,799 --> 00:10:04,799
PTSD or depression have this kind of brain. But it's

157
00:10:04,840 --> 00:10:06,519
only in the snapshow. We don't know if their brain

158
00:10:06,600 --> 00:10:09,200
was like that before. Did the depression cause it that

159
00:10:09,279 --> 00:10:12,600
the brain changes caused the depression, so reverse.

160
00:10:12,360 --> 00:10:13,759
Speaker 2: Causality could be an issue.

161
00:10:14,279 --> 00:10:18,080
Speaker 1: Another example of a cross sectionals comparing current anxiety levels

162
00:10:18,120 --> 00:10:20,000
between a group of teenagers and a group of adult

163
00:10:20,120 --> 00:10:25,080
older adults in May of twenty twenty six, so again one.

164
00:10:24,919 --> 00:10:25,679
Speaker 2: Point in time.

165
00:10:26,200 --> 00:10:29,679
Speaker 1: A longitudinal study designed that tracks the same group over time.

166
00:10:29,759 --> 00:10:34,080
It is uniquely capable of revealing development, stability, and causality.

167
00:10:34,639 --> 00:10:38,399
The down size is high risk of participant attrition, a

168
00:10:38,480 --> 00:10:41,159
lot of them drop out, high costs keeps a lot

169
00:10:41,200 --> 00:10:43,840
of the from ever getting started, and it also requires

170
00:10:43,840 --> 00:10:48,519
a massive time investment. Examples measuring and tracking changes and

171
00:10:48,559 --> 00:10:51,840
self esteem within the same exact group of individuals annually

172
00:10:52,440 --> 00:10:57,440
from ages thirteen through nineteen. Some other insights between subjects

173
00:10:57,480 --> 00:11:01,799
versus within subjects. Between subjects where different participants are assigned

174
00:11:01,840 --> 00:11:06,559
to each condition, the benefit completely eliminates carryover effects, which

175
00:11:06,600 --> 00:11:10,519
is where participating in one condition taints performance in the next.

176
00:11:11,279 --> 00:11:12,320
Speaker 2: A risk can.

177
00:11:12,279 --> 00:11:16,360
Speaker 1: Increase variability between groups due to individual differences. With thin

178
00:11:16,480 --> 00:11:20,799
subjects a design where the same participants experience all conditions,

179
00:11:21,679 --> 00:11:26,000
the benefit is that crucially reduces error due to individual differences.

180
00:11:26,039 --> 00:11:30,279
Because every participant acts as their own baseline control, there

181
00:11:30,320 --> 00:11:34,000
is a high risk of order effects, fatigue, practice, boredom.

182
00:11:34,200 --> 00:11:38,759
This risk is systematically solved with counterbalancing though randomizing or

183
00:11:38,799 --> 00:11:39,879
alternating the order.

184
00:11:39,720 --> 00:11:41,120
Speaker 2: In which conditions are presented.

185
00:11:41,559 --> 00:11:44,559
Speaker 1: Let me give you an example, testing a person's mood

186
00:11:44,600 --> 00:11:48,120
after listening to classical music versus metal music and a

187
00:11:48,159 --> 00:11:51,240
between groups design, you would have group A hearing classical

188
00:11:51,279 --> 00:11:55,080
and group B hearing metal. And within subjects design, everyone

189
00:11:55,120 --> 00:11:58,000
hears classical on day one, everyone on day two, here's

190
00:11:58,039 --> 00:12:03,000
metal and with half counterbalance to hear metal first. Others

191
00:12:03,000 --> 00:12:06,440
are factorial designs. These are research designs used to examine

192
00:12:06,519 --> 00:12:11,440
multiple independent variables and their interactions simultaneously. So how you

193
00:12:11,519 --> 00:12:16,639
expressed via numbers like two times two equals two independent

194
00:12:16,720 --> 00:12:20,960
variables each with two levels. It's excellent for examining how

195
00:12:21,039 --> 00:12:25,039
variables combine to influence psychological inco outcomes. On the E

196
00:12:25,159 --> 00:12:27,360
triple P, you must remember that both the main effects

197
00:12:27,360 --> 00:12:30,879
and the interaction effects should be interpreted carefully to understand

198
00:12:30,919 --> 00:12:36,919
how independent variables operate, both separately and jointly. Give me

199
00:12:36,919 --> 00:12:39,679
give you an example or study looking at whether treatment

200
00:12:39,720 --> 00:12:43,840
types CBT versus medication interacts with age group adolescent versus

201
00:12:43,879 --> 00:12:48,440
adult to affect depression scores. Heading over to validity trade

202
00:12:48,440 --> 00:12:52,679
offs in selection, So internal validity did the independent variable

203
00:12:52,799 --> 00:12:55,600
cause the change and the dependent variable? It gives you

204
00:12:55,720 --> 00:13:04,519
confidence and causality, but confounding variable selection bias, attrition history, maturation,

205
00:13:06,519 --> 00:13:11,159
external validity confinings generalize to other people in settings, sampling, bias,

206
00:13:11,320 --> 00:13:16,200
artificial settings, reactivity the core trade off as lab studies

207
00:13:16,240 --> 00:13:22,440
gained control, boosting internal validity, but may sacrifice generalizability, lowering

208
00:13:22,480 --> 00:13:27,279
external validity. Conversely, field studies studies do the opposite. So

209
00:13:27,320 --> 00:13:29,679
how do you match designs to research questions? Well, when

210
00:13:29,679 --> 00:13:33,399
you're choosing one, you always ask what's the question, what's feasible,

211
00:13:33,440 --> 00:13:35,480
what's ethical, and what's valid.

212
00:13:36,000 --> 00:13:37,600
Speaker 2: For instance, if you want to prove.

213
00:13:39,559 --> 00:13:41,240
Speaker 1: Want to prove cause, I guess this is a question

214
00:13:41,279 --> 00:13:45,120
you have to ask yourself. Experimental can't randomize, but need

215
00:13:45,200 --> 00:13:51,799
causal inference. All right, let's take a deeper dive into

216
00:13:51,879 --> 00:13:56,200
the validity section here. So validity refers to get to

217
00:13:56,240 --> 00:13:59,679
how accurate, trustworthy, meaningful the conclusions of a study are,

218
00:14:01,879 --> 00:14:06,480
and your researchers are always balancing control, realism, ethics, and generalizability,

219
00:14:06,519 --> 00:14:10,200
so internal validity. Some of the major threats for internal

220
00:14:10,279 --> 00:14:16,559
validity confounding variables, selection, bias, history, maturation, attrition, and testing effects.

221
00:14:17,639 --> 00:14:21,120
An example, did CBT actually reduce depression or patients also

222
00:14:21,240 --> 00:14:25,559
starting medication at the same time? For external validity, can

223
00:14:25,600 --> 00:14:28,080
the findings generalize to other people? Some of the major

224
00:14:28,120 --> 00:14:34,679
threats are sampling bias, artificial settings, participant reactivity, and narrow demographics.

225
00:14:35,440 --> 00:14:37,600
It laps that are using only college freshmen may not

226
00:14:37,639 --> 00:14:39,679
apply to elderly patients with depression.

227
00:14:40,360 --> 00:14:41,799
Speaker 2: What about construct validity?

228
00:14:41,799 --> 00:14:45,919
Speaker 1: Did the research or accurately measure the psychological concept they

229
00:14:45,960 --> 00:14:46,960
intended to measure.

230
00:14:47,679 --> 00:14:49,159
Speaker 2: Some of the threats include.

231
00:14:48,879 --> 00:14:54,320
Speaker 1: Poor operational definitions, measurement error, social desirability bias, and examples.

232
00:14:54,360 --> 00:15:00,440
Does a stress questionnaire actually measure stress or just general negativity.

233
00:15:01,360 --> 00:15:05,759
Another one is statistical conclusion validity, where the statistical conclusions

234
00:15:05,919 --> 00:15:11,639
accurate or appropriate. Some of the threats low power, unreliable measures,

235
00:15:11,720 --> 00:15:16,120
multiple comparisons, restricted range. A study may fail to detect

236
00:15:16,159 --> 00:15:19,000
the real effect because the sample size was just too small.

237
00:15:20,159 --> 00:15:23,320
Internal validity is the gold standard for causal inference. The

238
00:15:23,440 --> 00:15:27,000
higher the internal validity, the more confident researchers have that

239
00:15:27,120 --> 00:15:31,200
the independent variable cause the dependent variable, the findings are

240
00:15:31,240 --> 00:15:33,879
not due to chance, and alternative explanations have been ruled

241
00:15:33,879 --> 00:15:37,320
out to the best of their ability. High internal validity

242
00:15:37,360 --> 00:15:42,799
often requires random assignment, experimental control, standardized procedures, control groups,

243
00:15:42,799 --> 00:15:46,399
and elimination as the many confounds as possible. Some of

244
00:15:46,399 --> 00:15:49,240
the major threats were mentioned confounding variables and confounds an

245
00:15:49,279 --> 00:15:53,559
uncontrolled factor that changes alongside the independent variable and may

246
00:15:53,600 --> 00:15:57,320
explain the results. So an example, a psychologist studies whether

247
00:15:57,399 --> 00:16:02,600
meditation reduces anxiety. However, pticipants of the meditation group also

248
00:16:02,720 --> 00:16:04,600
begin exercising more frequently.

249
00:16:05,919 --> 00:16:07,559
Speaker 2: Now has become unclear.

250
00:16:07,679 --> 00:16:12,039
Speaker 1: Was anxiety reduced because of meditation, exercise or the combination.

251
00:16:13,120 --> 00:16:18,399
The exercise variable itself becomes a confound selection Bruse groups

252
00:16:18,440 --> 00:16:21,759
differ before the study begins. The school introduces a new

253
00:16:21,799 --> 00:16:26,360
reading intervention only in honors classrooms. Well, if students improve,

254
00:16:26,559 --> 00:16:31,919
the improvement may reflect higher baseline intelligence already, greater motivation,

255
00:16:32,080 --> 00:16:34,799
or family support, rather than the intervention itself.

256
00:16:34,919 --> 00:16:36,679
Speaker 2: That's why you need a more diverse group.

257
00:16:37,879 --> 00:16:41,679
Speaker 1: History effects, external events occurring during the study influence outcomes.

258
00:16:42,480 --> 00:16:47,000
For instance, researchers examine stress levels during a workplace intervention program,

259
00:16:47,000 --> 00:16:51,279
but halfway through the study the company announces layoffs. Stress

260
00:16:51,399 --> 00:16:57,559
changes may reflect the layoffs and not the intervention. Another one,

261
00:16:58,000 --> 00:17:04,319
participants naturally change over time. Children improve on emotional regulation

262
00:17:04,400 --> 00:17:07,640
after six months of therapy is an example, But children

263
00:17:07,720 --> 00:17:10,880
naturally mature emotionally over time anyway, making it difficult to

264
00:17:10,880 --> 00:17:15,559
isolate the treatment effect, and finally attrition participants drop out

265
00:17:15,599 --> 00:17:19,119
unevenly across groups. Example, and a substance abuse treatment study,

266
00:17:19,519 --> 00:17:23,599
the most severe participants leave treatment early. The remaining sample

267
00:17:23,680 --> 00:17:29,000
may falsely make the treatment appear more effective. External validity

268
00:17:29,039 --> 00:17:32,359
is another way of saying generalizability.

269
00:17:32,839 --> 00:17:34,319
Speaker 2: We talked about this earlier.

270
00:17:35,160 --> 00:17:38,240
Speaker 1: Researchers so as a study may have excellent internal validity

271
00:17:38,319 --> 00:17:42,240
but poor real world relevance, so researchers would ask would

272
00:17:42,240 --> 00:17:46,519
these findings apply in real clinical settings, different cultures.

273
00:17:46,039 --> 00:17:46,920
Speaker 2: Age groups, etc.

274
00:17:48,119 --> 00:17:50,799
Speaker 1: Some of the threats during external validity as sampling bias.

275
00:17:50,960 --> 00:17:54,799
The sample does not represent the broader population. A For instance,

276
00:17:54,839 --> 00:17:58,599
a study on burnout includes only affluent private practice psychologists.

277
00:17:59,160 --> 00:18:02,359
This one, in general may not generalize the community clinics, hospitals,

278
00:18:02,480 --> 00:18:07,160
or graduates. Artificial settings our laboratory environments may not reflect

279
00:18:07,240 --> 00:18:11,559
real life behavior, so participants may behave differently discussing trauma

280
00:18:11,680 --> 00:18:13,880
in a lab than they would in actual therapy sessions

281
00:18:14,559 --> 00:18:20,000
and reactivity participants alter behavior because they know they're being studied,

282
00:18:20,039 --> 00:18:23,920
sometimes called the Hawthorn effect. People may report eating healthier

283
00:18:24,039 --> 00:18:27,000
during nutrition research because they want to appear socially desirable

284
00:18:28,039 --> 00:18:31,599
construct validity examines whether the researcher truly measure the intended

285
00:18:31,599 --> 00:18:35,640
psychological construct. This is especially important in psychology because constructs

286
00:18:35,720 --> 00:18:40,720
like intelligence, depression, resilience cannot be directly observed. They must

287
00:18:40,759 --> 00:18:46,039
be operationalized through measurable indicators. Example of poor construct vibility

288
00:18:46,119 --> 00:18:49,000
is a research or claims to measure academic motivation by

289
00:18:49,039 --> 00:18:54,319
counting class attendances only, but attendants can reflect parental pressure,

290
00:18:54,880 --> 00:18:59,799
transportation access, or fear of punishment. The operational definition poorly

291
00:18:59,799 --> 00:19:04,960
can aptures the intended construct. What about statistical conclusion validity?

292
00:19:05,000 --> 00:19:08,160
This form of validity evaluates whether the correct statistics were used,

293
00:19:08,599 --> 00:19:11,359
the study had enough power, and the conclusions drawn are

294
00:19:11,440 --> 00:19:16,279
mathematically justified. Some of the common threats are low statistical power,

295
00:19:16,440 --> 00:19:19,880
small samples, increase the risk of type two errors, false negatives,

296
00:19:20,279 --> 00:19:23,720
missing real effects. A treatment truly works, but the sample

297
00:19:23,759 --> 00:19:28,440
size it's too small to detect the significance, unreliable measurement,

298
00:19:28,640 --> 00:19:33,000
unstable measures, weakened findings. Example, if a depresspression scale gives

299
00:19:33,039 --> 00:19:37,319
drastically different scores every few days, that's without real mood changes,

300
00:19:37,440 --> 00:19:43,119
the statistical conclusions become questionable. Internal versus external validity is

301
00:19:43,119 --> 00:19:45,000
one of the most heavily tested concepts on the E

302
00:19:45,039 --> 00:19:49,359
triple P the SEESAW principle. Lab studies high control, strong

303
00:19:50,519 --> 00:19:54,839
internal validity, lower realism, and lower external validity. Field studies

304
00:19:54,920 --> 00:20:00,000
high realism, strong external but lower internal validity and less control.

305
00:20:01,279 --> 00:20:02,440
Speaker 2: Give you another example.

306
00:20:02,519 --> 00:20:06,559
Speaker 1: A lab example, A memory researcher places participants in a

307
00:20:06,640 --> 00:20:11,440
silent control room with identical lighting and scripted instructions. Excellent control,

308
00:20:11,559 --> 00:20:16,039
few confounds, high internal validity, yet real world memory rarely

309
00:20:16,079 --> 00:20:21,880
occurs under such sterile conditions. All right, Matching research designs

310
00:20:21,920 --> 00:20:23,720
to research questions, here we go.

311
00:20:25,519 --> 00:20:27,480
Speaker 2: Actually, I forgot to tell you about field example.

312
00:20:27,559 --> 00:20:32,839
Speaker 1: Psychology studies police decision making during actual patrol shifts advantages

313
00:20:32,920 --> 00:20:41,000
realistic behavior and stronger ecological validity, disadvantages uncontrolled variables, environmental unpredictability,

314
00:20:41,079 --> 00:20:48,039
and low internal validity.

315
00:20:49,400 --> 00:20:57,359
Speaker 2: All right, So.

316
00:20:57,359 --> 00:21:02,279
Speaker 1: It's where recap experimental design establishes calledation quasi experimental approximate

317
00:21:02,319 --> 00:21:08,599
causation without randomization exploring associations or predictions is correlational examine

318
00:21:08,680 --> 00:21:13,440
changes across time, logitudin and study rare or highly complex phenomena.

319
00:21:13,440 --> 00:21:18,319
As a case study, validity often competes also with other ethics,

320
00:21:18,319 --> 00:21:21,599
Some studies with the highest possible internal validity would be unethical.

321
00:21:22,000 --> 00:21:26,920
Researchers cannot randomly assign children to abuse and neglect or trauma. Therefore,

322
00:21:26,960 --> 00:21:32,079
psychologists often rely on quasi experiment longitudinal correlational studies. These

323
00:21:32,119 --> 00:21:36,680
desize designs improve ethics, but we can causal certainty. Here

324
00:21:36,680 --> 00:21:39,359
are some practice questions for you. First one, a researcher

325
00:21:40,119 --> 00:21:44,160
is studying a rare neurological condition and follows three individuals

326
00:21:44,160 --> 00:21:49,880
over five years using behavioral observations, interviews, and neural imaging.

327
00:21:50,000 --> 00:21:58,640
What is the primary primary methodological limitation? If you said

328
00:21:58,680 --> 00:22:04,400
B I mean external validity, you're right. Case studies provide

329
00:22:04,440 --> 00:22:08,400
exceptionally rich and detailed information, especially for rare conditions. However,

330
00:22:09,160 --> 00:22:13,359
findings from only three individuals cannot generalize to larger populations.

331
00:22:14,759 --> 00:22:18,640
Case studying methodology is a weakness here qualitative death and

332
00:22:18,720 --> 00:22:20,119
highly specialized samples.

333
00:22:20,680 --> 00:22:21,720
Speaker 2: Question number two.

334
00:22:22,240 --> 00:22:27,480
Speaker 1: A psychologist measures employee stress levels before wellness intervention, midway through,

335
00:22:31,599 --> 00:22:32,680
immediately after.

336
00:22:32,480 --> 00:22:33,599
Speaker 2: And six months later.

337
00:22:33,799 --> 00:22:36,319
Speaker 1: The design is best classified as what is it a

338
00:22:36,400 --> 00:22:42,240
factorial design, cross sectional or time series design. If you

339
00:22:42,279 --> 00:22:46,200
said time series design, you're right. Time series designs repeatedly

340
00:22:46,240 --> 00:22:51,160
measure the same variable across multiple time points. These designs

341
00:22:51,160 --> 00:22:55,640
are especially useful when randomization is impossible, interventions occur naturally,

342
00:22:55,759 --> 00:23:00,240
or researchers want to examine trends. Question number three two

343
00:23:00,319 --> 00:23:04,920
by three factorial design studying anxiety treatment. How many independent

344
00:23:05,000 --> 00:23:07,400
variables in total conditions exist?

345
00:23:09,039 --> 00:23:10,319
Speaker 2: Is it A two.

346
00:23:10,160 --> 00:23:14,920
Speaker 1: Independent variables in five conditions? Does it B two independent

347
00:23:15,000 --> 00:23:21,200
variables in six conditions or three independent variables and two conditions.

348
00:23:24,079 --> 00:23:27,480
If you said B two independent and six conditions, you're right.

349
00:23:27,559 --> 00:23:30,759
The number of digits equals the number of ivs independent

350
00:23:30,799 --> 00:23:34,279
variables is two times three is six, and there you go.

351
00:23:35,559 --> 00:23:36,400
Speaker 2: Question number four.

352
00:23:36,559 --> 00:23:40,200
Speaker 1: A developmental psychologist wants to reduce error variants caused by

353
00:23:40,200 --> 00:23:44,519
participant differences while testing three educational apps.

354
00:23:44,559 --> 00:23:46,000
Speaker 2: Which design is best.

355
00:23:46,559 --> 00:23:52,839
Speaker 1: Between subjects, cross sectional or within subjects design with counterbalancing?

356
00:23:54,359 --> 00:23:57,440
Do you want to reduce error variants? If you say

357
00:23:57,440 --> 00:24:02,200
within subjects, you're right. Within subjects produce error because each

358
00:24:02,240 --> 00:24:06,079
participant acts as their own control. However, order effects become

359
00:24:06,119 --> 00:24:09,960
a major concern, and counterbalancing can help reduce fatigue and

360
00:24:10,000 --> 00:24:14,079
practice effects. And the last question, a study finds a

361
00:24:14,079 --> 00:24:18,880
correlation between clinician experience and patient retention. Credit Critics argue

362
00:24:18,920 --> 00:24:21,880
older clinicians tend to work in a wealthier neighborhoods, which

363
00:24:21,920 --> 00:24:27,079
may explain retention outcomes. The criticism reflects what attrition bias,

364
00:24:28,359 --> 00:24:35,119
the third variable problem or testing effects. If you said

365
00:24:35,160 --> 00:24:38,920
the third variable problem, you're right. An uncontrolled third variable

366
00:24:38,960 --> 00:24:43,720
problem like a wealthy neighborhood may explain placement and repatient retention.

367
00:24:44,200 --> 00:24:50,200
Correlation is not equal causation. Some the triple P study

368
00:24:50,240 --> 00:24:54,000
tips think clinically, not definitionally. The E triple P rarely

369
00:24:54,039 --> 00:24:56,599
asks what is internal validity instead of presents a videt

370
00:24:56,680 --> 00:24:59,319
and asks what threat exists, what designers.

371
00:24:58,920 --> 00:25:00,720
Speaker 2: Used, or what conclusion is justified.

372
00:25:01,799 --> 00:25:05,119
Speaker 1: Immediately identify the independent and dependent variables and ask yourself

373
00:25:05,160 --> 00:25:06,039
what is manipulated?

374
00:25:06,039 --> 00:25:09,160
Speaker 2: What is measured? No manipulation no, No.

375
00:25:09,279 --> 00:25:16,759
Speaker 1: Random assignment usually means correlational, observational and quasi experimental. Learn

376
00:25:16,799 --> 00:25:20,839
the validity language external validity generalized to real life, eternal,

377
00:25:21,079 --> 00:25:27,839
alternative explanation, construct accurately measures and then statistical sample too small.

378
00:25:28,640 --> 00:25:31,599
Remember the lab versus steady trade offs we talked about.

379
00:25:31,759 --> 00:25:37,039
Know why correlation cannot prove causation. Three major reasons directionality,

380
00:25:37,240 --> 00:25:41,279
third variable or confounds, and lack of manipulation and random assignment.

381
00:25:41,680 --> 00:25:47,400
The E triple P repeated tests these concepts in disguised forms. Finally,

382
00:25:47,680 --> 00:25:51,400
here's a high yield takeaway when evaluating any research study.

383
00:25:51,480 --> 00:25:54,680
Train yourself to ask four questions. Can I trust the

384
00:25:54,720 --> 00:26:00,440
cause of conclusion? Internal validity, Can this generalize external ability?

385
00:26:01,039 --> 00:26:04,079
Did they measure the right construct construct ability? And where

386
00:26:04,119 --> 00:26:06,720
the statistical conclusions justified?

387
00:26:07,920 --> 00:26:11,039
Speaker 2: Well, there you go, folks. Hopefully this is a little longer.

388
00:26:11,160 --> 00:26:13,359
Speaker 1: Than I expected on the podcast, and hopefully you didn't

389
00:26:13,400 --> 00:26:15,759
mind that. Hopefully you got a lot of insight. But

390
00:26:15,799 --> 00:26:18,880
you know, research method statistics. A lot of people find

391
00:26:18,880 --> 00:26:21,880
this difficult in psychology programs, not what they're used to,

392
00:26:23,079 --> 00:26:23,759
So I get it.

393
00:26:24,960 --> 00:26:26,039
Speaker 2: That's it for now, folks,